October 1-5, 2018; Berlin, Germany; Full Report – Draft

Executive Highlights

In this report, we provide our full coverage of diabetes/obesity therapy, diabetes technology, and big picture topics from EASD 2018. There were 15,699 attendees this year. Dive into 11 chapters, ranging from GLP-1s and SGLT-2s, to diabetes technology and novel therapies. Talk titles highlighted in yellow denote the presentations and commentary that we found particularly notable. Titles highlighted in blue represent new or expanded coverage that wasn’t included in our daily highlights (day #1, day #2, day #3, day #4, and day #5). 

Note that some talks may appear in multiple sections (e.g. content on basal insulin/GLP-1 fixed-ratio combinations appears under the “GLP-1” category as well as the “insulin” category). Navigate through all 185 pages of reporting using our table of contents below, starting with our top 12 key themes from the meeting. Also, many of the videos from the sessions are available online at this link – happy reading!

Themes

Diabetes Therapy

1. GLP-1 Agonists: A Heterogenous Class Tethered by a Common Thread of Cardioprotection? New Evidence Supports CV Class Effect

• New data and analyses released at EASD 2018 offer insight on lingering questions about the cardioprotective effects of GLP-1 agonists. In a very unexpected finding, the HARMONY-Outcomes CVOT for GSK’s now-discontinued GLP-1 Tanzeum (once-weekly albiglutide) achieved significance in reducing risk of three-point MACE vs. placebo, with an impressive 22% relative risk reduction (HR=0.78, 95% CI: 0.68-0.90, p<0.0001 for non-inferiority, p=0.0006 for superiority). Heightening the intrigue accompanying this positive result was a whirlwind of factors surrounding Tanzeum: GSK pulled the drug from market over the past year – probably unsurprising given global pricing pressure (it was explained by terminology of “a restructuring”), and albiglutide itself is widely regarded as offering a clinical profile with less weight loss relative to other GLP-1s and less A1c reduction. In HARMONY, the A1c reductions not as impressive as with other compounds (~0.6% treatment difference at eight months), weight loss was not meaningful (0.66 kg difference at eight months), and differences in blood pressure were also not seen. In spite of these uninspiring results on traditional markers for diabetes therapies, albiglutide still delivered a significant reduction in MACE events, driven by a 25% reduction on MI (HR=0.75, 95% CI: 0.61-0.90). This is a big deal. As Dr. John Buse emphasized to us, HARMONY offers strong evidence that GLP-1 driven cardioprotection is mediated through the GLP-1 receptor, rather than through effects on glycemia or weight. All-in-all, while the field was already trending toward consensus on a class effect for GLP-1 agonists, HARMONY offers another strong and critical piece of evidence in support of this notion, though questions certainly remain as to which population those effects apply to.

• The heterogeneity of GLP-1 CVOT design offers a compelling explanation for variable results seen to date. As Dr. John Buse put it: “If you want to win a cardiovascular outcomes trial, start with people with clinical cardiovascular disease and an A1c above 8%.” HARMONY followed this blueprint to a tee, enrolling what Dr. David Matthews called a “really, really sick” population, with 100% of participants having established CVD, poor glycemic control, and heavy baseline medication use. As Dr. Matthews explained, these patients were on the brink of having a MACE event but far enough from the edge that albiglutide could still demonstrate a positive effect. By enrolling this population, HARMONY met superiority within a median follow-up of only 1.6 years – the shortest of any GLP-1 CVOT to date. When compared to the more pragmatic design of EXSCEL, the importance of trial design becomes readily apparent – and Dr. Buse presented a post hoc analysis demonstrating just that. EXSCEL missed statistical superiority on three-point MACE by the most razor-thin of margins (HR=0.91, 95% CI: 0.83-1.00, p=0.0061), and has been a lightning rod for discussion on GLP-1 cardioprotection and the importance of trial design. But, as presented by Dr. Buse, censoring the use of open-label drop-in medications (including protocol-violating concomitant GLP-1 prescriptions) shifted EXSCEL results to superiority on three-point MACE. When accounting for only sulfonylurea drop-ins, the HR on three-point MACE shifted to nominal superiority (HR=0.98, 95% CI: 0.81-0.99, p=0.024); a more pronounced effect was seen when accounting for all drop-ins (HR=0.87, 95% CI: 0.78-0.97, p=0.011). This analysis fits with the broader picture we’ve heard from numerous KOLs since EXSCEL results were first released: Bydureon most likely confers cardioprotection, but aspects of trial design – including the lack of a run-in period to exclude low-adherence, a high proportion of primary prevention patients, and use of complicated reconstitution kits – worked against the GLP-1 in EXSCEL.

• Despite strong evidence that GLP-1 cardioprotection occurs independent of glucose-lowering and weight loss, there’s no doubt that effects on glucose and weight are of very high importance. As some have pointed out, better effects on reducing hyperglycemia and driving meaningful weight loss could very well translate to a more substantial reducing in longer-term CV risk. In the shorter-term, though, GLP-1 receptor-mediated anti-atherosclerotic effects seem most meaningful. In HARMONY, the progressive and steady separation of Kaplan-Meier curves indicates an anti-atherosclerotic effect, in line with the effects seem in LEADER, SUSTAIN-6, and somewhat in EXSCEL. This phenomenon is all the more obvious when compared SGLT-2 inhibitor CVOTs, where a sharp initial separation of curves (within mere weeks) is followed by a lack of progressive separation; this trend is in line with the hemodynamic effects most now think drive the CV benefit associated with SGLT-2s. In their reactions to HARMONY results, Drs. Naveed Sattar and Darren McGuire both pointed to the modest efficacy of albiglutide in reducing weight and blood glucose as further evidence for the putative anti-atherosclerotic mechanism through which GLP-1s are thought to confer cardioprotection. While our sense is that very few thought A1c lowering or weight loss played a substantial role in the effects seen in LEADER or SUSTAIN 6, the point stands that these agents offer more impressive A1c lowering and weight loss than Tanzeum. As such, HARMONY has provided the strongest evidence yet supporting the independence of cardioprotection and glucose-lowering with GLP-1s, at least in the short-term follow-up of the trial. This being said, the “legacy” benefits of better glucose lowering and weight loss should not be forgotten. We would imagine that over a longer-term follow-up, more potent and sustained effects on these endpoints could translate to a higher magnitude of CV risk reduction – we also imagine the cost of demonstrating this would be prohibitive. To this end, HARMONY commentator Dr. David Matthews postulated that albiglutide could be most useful in normal and underweight patients with diabetes, including those with type 1 (though of course many people with type 1 would benefit from weight loss as well).

• The heterogeneity of GLP-1 agonist molecular structures still fuels some debate. Thus far, the only trials to demonstrate definitive cardioprotection have tested molecules with high homology to human GLP-1 (HARMONY with albiglutide, LEADER with liraglutide, and SUSTAIN-6 with semaglutide). Meanwhile, the two CVOTs to miss superiority have tested molecules based on exendin-4 (ELIXA for lixisenatide and EXSCEL for exenatide). However, in both cases, there are critical aspects of trial design and enrolled population that are often cited as meaningful hurdles to achieving superiority, an argument that has been vocally and widely expressed for EXSCEL in particular – speakers often group EXSCEL in with “positive” trials. We often hear thought leaders cite this difference as one possible explanation of differing results within the class; for example, Dr. Daniel Drucker has been a strong advocate for this position in the past. That said, our sense is that the field is warming up to the idea of cardioprotection as a defining characteristic of the entire GLP-1 class, regardless of molecular composition. While differences in molecular structures may contribute to variations in signaling and even some clinical effects, we understand it’s more likely that varying trial designs are the primary factor contributing to heterogeneity in GLP-1 CVOT outcomes.

  • It does, however, seem more likely that a GLP-1 agonist’s duration of action could be important to CV effects. With the addition of HARMONY data and consideration of post-hoc analyses of EXSCEL, every long-acting GLP-1 (liraglutide, exenatide, semaglutide, and albiglutide) arguably offers cardioprotection. The lone exception in the GLP-1 class, so far, is Sanofi’s Adlyxin (lixisenatide), which in the ELIXA trial demonstrated resoundingly neutral CV effects. We’ve heard KOLs attribute these results to the short-acting nature of lixisenatide (two to four hours, compared to ~12 hours for liraglutide, ~seven days for semaglutide, ~two weeks for exenatide XR, and ~five days for albiglutide). Perhaps, as Dr. Lars Ryden commented, the bioavailability of the GLP-1 agonist must last at least through one day in order to confer a reduction in CV risk. The REWIND CVOT for Lilly’s once-weekly Trulicity should speak further to this question (results expected by end of 2018), but we’ve also heard commentary that the entirely post-recent-ACS population enrolled in ELIXA was likely too sick for lixisenatide to confer any benefit (i.e., the population was ‘too far gone’). Regardless of such speculation, the upshot for patients is a clear win: Multiple GLP-1s with demonstrated CV protection are on the market, and we hope that albiglutide finds its way back into patients hands (perhaps even at a reduced price, though we understand the albumin-bound molecule is quite costly to produce!).

2. The Use of SGLT Inhibitors in Type 1 Diabetes Is More Complex Than Ever: EASE Results for Jardiance in Type 1 Add a New Layer to Risk/Benefit; Patient Satisfaction Highlighted

• Building on the strong emphasis seen at ADA 2018 just a few months ago, the use of SGLT inhibitors in type 1 diabetes continued to serve as a serious point of discussion at EASD 2018. A full SGLT-centric symposia focusing on type 1 adjunct therapy kicked off the meeting, and the following days delivered more intriguing discussions and data on the topic: Of particularly high note, full phase 3 EASE results for Jardiance in type 1 were presented on Thursday, and a notable Lexicon poster highlighted high patient satisfaction with sotagliflozin.

• EASE results for Jardiance in type 1 further complicate our understanding of the risk/benefit trade-off with SGLTs in type 1. Very notably, the 2.5 mg dose of empagliflozin was not associated with an increased risk of DKA in the way that the 10 mg and 25 mg doses were. But while this dose still gave significant improvements in time-in-range and A1c, the magnitude of those benefits was meaningfully lower than with the higher doses. This finding adds another layer of consideration to the risk/burden and benefit profile of the class. These include glucose lowering, DKA risk, reduced hypoglycemia frequency, weight loss, time-in-range benefits, psychosocial impacts, potential cardio (and renal) protection, and ketone monitoring burden – and all of these factors converge into a highly personal decision for each person with type 1 diabetes who might consider taking an SGLT inhibitor. To this end, we think the lower dose of empagliflozin could have an interesting role to play: Some patients and providers might prefer a somewhat less efficacious therapy if it comes with little to no increased DKA risk. Still, significant questions remain: How will regulatory agencies consider this lower dose relative to previously tested doses, which carry increased DKA risk? We’ll get our first taste of how FDA is approaching the issue when sotagliflozin is discussed at an expected FDA Advisory Committee in early 2019 (January 17).

• The benefits of SGLT inhibitors in meeting an unmet need in type 1 diabetes treatment is more apparent than ever, and data from Lexicon/Sanofi reflect high patient satisfaction with sotagliflozin. Dr. Thomas Danne highlighted this point, focusing on how SGLT inhibitors (and potentially other adjunct therapies) can help to relieve the enormous emotional burden associated with type 1 diabetes management. As the beyond-A1c-movement marches on, it’s clear that considering adjunct treatments will become more and more essential. Importantly, SGLT inhibitors not only improve A1c; they also increase time-in-range and relieve some of the stress and emotional burden tied to management of type 1 diabetes. On this front, a Lexicon-sponsored poster shared findings from a phone survey of those enrolled in phase 3 inTandem studies after study completion, offering strong patient support for the benefits conferred by SGLTs. Patients using the treatment explained how the greater glycemic stability offered by sotagliflozin yielded an improved sense of well-being and reduced levels of burden and stress. When considering DKA risk, it’s important to keep this point in mind: There is a huge need for better treatment options in type 1 diabetes, and any option that can help fill this area of unmet need must be seriously considered. As well, there is major opportunity to avoid DKA by having a very structured process for how to address dangers.

• DKA risk remains at the core of the conversation. Any conversation or presentation on SGLT use in type 1 is underscored by the small but significant increased risk for DKA conferred by these agents. EASD 2018 gave us a lot to think about regarding this problem, particularly in light of EASE results, which demonstrated a dose-dependent response on both efficacy and safety. Most importantly, an increased risk of DKA occurred in the 10 mg and 25 mg empagliflozin groups, but not in the lower 2.5 mg dose group – while promising, we note that efficacy also meaningfully fell off in the lower-dose group. In EASE, intensive ketone monitoring was recommended to patients, specifically daily blood betahydroxybutyrate measurements to establish baseline levels during the first four weeks of treatment and thereafter two to three measurements per week. Even in light of these quite stringent ketone monitoring guidelines, DKA events still occurred at elevated rates for the higher doses of empagliflozin. In our view, the fact that DKA events occurred at an increased rate despite intensive ketone monitoring solidifies the reality of DKA as a very real and important risk to consider with adjunct SGLT use in type 1 diabetes. How thoroughly patients actually adhered to ketone monitoring guidelines is currently uncertain, and it’s very certain (to us) that suboptimal monitoring may very well contribute to elevated DKA risk. However, better patient and provider education around DKA risk and the importance of ketone monitoring can help to mitigate these adverse events, and it will take an integrated effort from to tailor general and personalized strategies for minimizing DKA risk.

• Calls for a CVOT in type 1 diabetes are more prominent than ever. We heard multiple thought leaders – including Dr. Julio Rosenstock and Dr. Thomas Pieber – advocate for an SGLT CVOT in type 1s, and we couldn’t agree more with this sentiment. It has been emphasized that those with type 1 are drawn to adjunct SGLT use not only for the A1c and time-in-range benefits, but also for potential organ (heart and kidney) protective effects that have been established in type 2. After all, increased cardiovascular and renal risk is not unique to type 2 diabetes; those with type 1 also experience excess risk of CV events and, of course, microvascular complications. We thought Dr. Rosenstock captured this idea best on day one of EASD 2018: “If we’re all so excited about SGLTs for CV disease, heart failure, and nephropathy in type 2 diabetes, we need to honestly ask ourselves if these effects can be replicated in type 1.” Mechanistically, there’s reason to believe these effects of SGLT inhibition could play out similarly in type 1. However, hard evidence in the way of a completed CVOT is surely needed to promote broad use of these agents for reducing risk. We expect to get a small glimpse into these potential effects through ongoing studies of these agents that aren’t excluding patients with type 1 diabetes, including Lilly/BI’s EMPA-KIDNEY and EMPEROR-HF studies of Jardiance in CKD and heart failure, respectively. To be sure, challenges of funding, enrollment, and trial design present sizable challenges to a potential type 1 CVOT, but the insight to be gained through such a study would is immensely important.  

3. Dual Agonists Break Through with Unprecedented Weight Loss and A1c Lowering in Phase 2; Lilly’s Tirzepatide to Become First-Ever Phase 3 Dual Agonist, Entering SURPASS Program in 2019

• EASD 2018 certainly marked an inflection point for dual agonists, as Lilly released highly-anticipated and unprecedented results for its GLP-1/GIP dual agonist LY3298176 (now named tirzepatide), and MedImmune/AZ put forth similarly exciting phase 1/2 and phase 2a data for their GLP-1/glucagon dual agonist MEDI0382. Based on these results, both molecules represent a potential huge leap forward in incretin therapeutics, diabetes treatment, and even obesity pharmacotherapy – the hype for dual agonist is now backed by strong clinical data. That said, significant and meaningful tolerability concerns (specifically with regard to GI side effects) have emerged as a potential Achilles heel for dual agonists, though Lilly’s dose-ranging study does indicate a balance can be struck between very strong efficacy and tolerability comparable to current GLP-1 agonists (titration protocol is another consideration – more on this below). Moreover, the definitiveness of our understanding of any given agent’s profile can only be so strong based on a single phase 2 study. Overall, it seems that dual agonists may well carry a greater burden of GI side effects, but it’s also important to balance this against very impressive efficacy on both glucose lowering and weight loss.

  • Phase 2b data for Lilly’s once-weekly GLP-1/GIP dual agonist LY3298176 (“LY”) – now called tirzepatide – astounded a packed audience: The highest 15 mg dose gave a 2.4% reduction in A1c and ~25 pounds of weight loss in just 26 weeks (n=316). Of those randomized to 15 mg, one-quarter achieved ≥15% weight loss and 30% achieved an A1c <5.7%, both of which were significant vs. placebo and Trulicity 1.5 mg. However, one-quarter in that group also discontinued treatment due to adverse events, and 35% discontinued treatment overall; 66% of the group experienced GI side effects. That said, optimism for the candidate (even at the highest dose) abounded for three reasons: (i) All of the lower doses of LY (1, 5, and 10 mg) saw similar rates of treatment discontinuation (4%, 9%, and 6%, respectively) to Trulicity (11%); (ii) the 10 mg dose conferred ~80% of the weight loss and A1c lowering of the 15 mg dose; and (iii) the negatively tolerability profile of the 15 mg dose was attributed to very rapid titration (over only six weeks). We do note that 50% of the LY 10 mg group experienced any GI side effects (vs. 43% with Trulicity 1.5 mg and 33% with LY 5 mg), which isn’t a very sizable increase over Trulicity but may present some additional challenge vs. Trulicity in real-world adherence, where patients lack the extensive support offered in clinical trials. On the other hand, Lilly management noted in a subsequent conference call that a slower, step-wise titration protocol tested in a smaller 12-week dosing study resulted in discontinuation rates below 5% with the 15 mg dose. Importantly, this will inform dosing protocol for Lilly’s eight-trial, phase 3 SURPASS program, which includes a comparative study vs. semaglutide and a superiority-powered CVOT. The program is set to begin in 2019, enabling a hopeful 2022 submission.

  • Posting similarly positive results, MedImmune and AZ’s GLP-1/glucagon dual-agonist MEDI0382 conferred a 22% drop in glucose AUC, ~7.4 pounds of weight loss, and a 25% increase in time-in-range vs. placebo in just seven weeks, per phase 2a results (n=65). Promisingly, weight loss had not plateaued at the end of the study; assuming that MEDI0382 continues along this weight-loss trajectory (a huge assumption, granted), the candidate could very well give a similar mean weight loss as that seen with Lilly’s tirzepatide. Similar to tirzepatide, GI tolerability was a meaningful issue for MEDI0382, as ~80% of those randomized to the therapy experienced a treatment-emergent adverse event – the vast majority of which were mild to moderate nausea and vomiting. Importantly, longer titration within this study did not significantly improve the candidate’s tolerability profile (though there’s only so much flexibility with a seven-week study). Indeed, in a phase 1/2 dosing study also released at EASD, treatment-related adverse events also occurred more often with MEDI0382 than with placebo (86% vs. 47%). To be sure, though, we’re excited for further investigation of this candidate: A much larger phase 2b dose-ranging trial (n=834) of MEDI0382 is expected to read out in 1H19; according to AZ’s 2Q18 update, this study will enable management to make a go/no-go decision on phase 3. Excitingly, the candidate is being considered for multiple indications in type 2 diabetes and NASH, and no fewer than six phase 2 investigations are currently ongoing. We’re particularly intrigued by a study of MEDI0382 in combination with SGLT-2 inhibitor dapagliflozin (expected completion December 2018).

    • On a broader level, we were thrilled to see CGM incorporated into the phase 2a trial – a decision which MedImmune has assessed as overwhelmingly positive on account of the fine-detail glucose effects revealed in analysis. For example, Dr. Victoria Parker explained that the glycemic benefits and reduced variability offered by MEDI0382 seem to begin within hours of the first dose, as demonstrated by decreased mean glucose levels immediately following the first administration of the dual agonist in CGM tracings.

• Wrapping up multi-agonist discussions at EASD was Dr. Matthias Tschöp, who reflected on this impressive new data and discussed the remarkable potential of tri-agonists. According to Dr. Tschöp, the preclinical efficacy of dual agonists translated remarkably well to clinical trials. If this principle holds for tri-agonists in the pipeline, it would bode extremely well for this next-next-generation of incretins, which have demonstrated an additive effect on weight loss and glucose lowering in preclinical studies, compared even to dual agonists. In this excitement, Dr. Tschöp shared a list of all co- and tri-agonists currently in development (slightly updated from this recent paper) which includes 12 GLP-1/glucagon dual-agonists, five GLP-1/GIP dual-agonists, and three GLP-1/glucagon/GIP tri-agonists. To conclude his presentation, Dr. Tschöp provided idyllic projections for a future of “personalized precision medicines” for obesity and type 2 diabetes. As he sees it, a variety of dual- and tri- agonists have the potential to allow greater responsiveness to medication based on better individualization to patient needs and medical history, and we’d also love to see precision medicine enable prediction of how individual patients will tolerate specific agonists. Given that GI side effects remain a meaningful barrier to adherence with GLP-1 agonists (and likely, in the future, dual agonists), we see great value in work aimed at improving patient experiences with tolerability in addition to efficacy. While we understand this is a relatively far ways away, the prospect is undeniably exciting.

4. Emphasis on Patient-Reported and Patient-Centric Outcomes; Victoza, Xultophy, and Sotagliflozin in Type 1 Demonstrate Satisfaction and/or Convenience

• Across diabetes treatment areas, EASD 2018 was rich with compelling patient-centered and patient-reported outcomes, both during oral presentations/symposia and in the poster hall. On a macroscopic scale, we were unsurprised by the results: Newer, more efficacious therapies, in general, improved patient satisfaction and experiences. In many ways, these results confirm that diabetes therapies are moving in the right direction and working to fill unmet needs for patients with both type 1 and type 2 diabetes. Moreover, we do think that drug efficacy and being able to see meaningful benefit is important for patient adherence, and these presentations serve as a reminder that anything HCPs can do to reinforce efficacy (e.g. showing patients data on how their meds are helping) can help reinforce adherence. While reducing long-term microvascular and macrovascular complications is the cornerstone of diabetes care, context is critical in patient care – some patients will be compelled by “XX% relative risk reduction on three-point MACE with a p-value of Y,” while others will be driven by “patients typically lose about X pounds when they start taking this drug.” We do hope that more relevant, patient-centered statistics such as these can become common components of data collection – and also be leveraged to improve patient engagement and excitement about treatment regimens.

  • GLP-1 agonists: Dr. Melanie Davies gave an illuminating talk demonstrating high patient satisfaction with GLP-1s, in particular noting improved satisfaction with high-dose Victoza vs. Januvia (p<0.05) and high-dose Victoza vs. Bydureon (p<0.0001) over the course of the 26-week LEAD-2 and LEAD-6 studies, respectively. These findings were actually published years ago by Dr. Davies, reflecting the relative dearth of patient satisfaction data but nonetheless supporting the notion that HCPs cannot predict which medicines patients will prefer. That is, many would expect an oral pill with few sides effects (DPP-4s) to be preferred to injectables with noted side effects (GLP-1s), but that’s not necessarily the case. We would love to see use of the Diabetes Treatment Satisfaction Questionnaire (DTSQ) standardized across clinical trials, as we think it could serve as an informative metric for HCPs in selecting medications in conversation with patients. As alluded to above, however, it’s important to note that medication preference is a very individualized subject; it’s not just about collecting one overarching “patient perspective,” but about appreciating each individual’s patient perspective. For many, a trade-off in tolerability, dose administration, or frequency of dosing is acceptable if it means improved efficacy. Summed up by Dr. Davies: “Treatment satisfaction is the whole package: It’s efficacy and tolerability, but it’s also the device and experience […] We can’t assume a patient is less satisfied with an injection than an oral, and we can’t assume that a daily medication is worse than a weekly one.”

  • Fixed-ratio basal insulin/GLP-1 agonist combinations: We often reference the impressive convenience, efficacy, and safety offered by basal/GLP-1 combinations, and a new DUAL VII analysis offered highly compelling evidence in favor of the patient-experience benefits of Novo Nordisk’s Xultophy (insulin degludec/liraglutide). According to Dr. Eden Miller, Xultophy treatment was less burdensome to patients than basal-bolus therapy in terms of dose adjustments, injection burden, and total daily insulin dose; as a reminder, DUAL VII compared Xultophy to basal-bolus treatment. While basal insulin adjustments were similar between the two groups (~17), the number of bolus insulin adjustments (inherently only in the basal-bolus group) averaged around 200 over the course of the 26-week trial. Moreover, the number of injections per day is always only one with Xultophy, compared to two to five for the basal-bolus group (an SMBG test was also required for each injection, adding extra burden to the basal-bolus group). Finally, total daily insulin dose at 26 weeks clocked in at a mean of ~40 units for the Xultophy group, compared to  an average of ~84 units for the basal-bolus group, insulin bolus insulin (p<0.0001); basal dose alone was also higher in the basal-bolus group. As a reminder, Xultophy demonstrated non-inferiority to basal-bolus on A1c lowering in DUAL VII, but also offered superiority on secondary outcomes including severe or symptomatic hypoglycemia (89% reduction) and weight change (-2.1 lbs with Xultophy vs. +5.8 lbs with basal/bolus). Of course, it’s important to note that no direct measure of patient satisfaction or quality of life was gathered, and we cannot assume that convenience necessarily translates to satisfaction for all patients (see Dr. Davies’ quote above). That said, we often hear thought leaders decry the complexity – and often, corresponding lack of efficacy – associated with basal-bolus regimens, and more needs to be done to make sure patients have information about and access to treatments, like Xultophy, that could help make their diabetes management easier.

  • SGLT inhibitors in type 1: A fascinating Lexicon poster detailed the patient perspective on SGLT-1/2 dual inhibitor sotagliflozin through post-study (inTandem1 or inTandem2) phone interviews with 40 study participants. Most notably, interviewees in the treatment arm of inTandem (n=27) saw diabetes treatment satisfaction rise from a mean of 3.9 to a striking 5.5 (on a 0-6 scale) while enrolled in the study, vs. a decrease in treatment satisfaction over the course of the study (from 4.3 to 3.7) for those on placebo (n=14). At a more granular level, the reasoning for this ringing endorsement for sotagliflozin was clear: 81% of sotagliflozin-treated participants noted fewer high blood sugar events and improved blood sugar stability (vs. 36% of interviewees on placebo), 48% reported that their insulin seems to work better (vs. 21% on placebo), and 37% reported fewer low blood sugar events (vs. 14% on placebo). Also to this point, a tremendous 89% of participants randomized to sotagliflozin stated that they would have liked to remain on the study medication (n=24), vs. only 43% of interviewees on placebo (n=6). It’s clear that patients noticed improvement while on sotagliflozin, and most were highly pleased with the drug’s effects. These sentiments were verbalized in direct patient testimony, which we highly encourage you to check out. Of course, there is likely a high degree of self-selection bias in these phone interviews. But perhaps most importantly in the context of Lexicon’s anticipated FDA Ad Comm, patients did not seem to be overburdened by inTandem’s ketone monitoring requirement, though to our understanding regular ketone checks were not recommended in the official protocol. Of the nearly half of interviewees reporting undergoing at least occasional ketone monitoring both before and during the trial, all reported believing that they could avoid DKA by both monitoring ketones and understanding the signs and symptoms of hyperglycemia. We certainly hope data such as this is at least mentioned during the Ad Comm for sotagliflozin in type 1 on January 17, as future treatment/safety recommendations will undoubtedly involve ketone monitoring in some form.

5. New CVOT Data Carries Significant Class Implications: Solidifies GLP-1 Cardioprotection As Class Effect Independent of Glycemic Control; Reassuring DPP-4/Heart Failure Data; Need for Better Obesity Pharmacotherapy

• With two readouts, EASD certainly did not lack in major CVOT data: We saw results from HARMONY for GSK’s discontinued GLP-1 agonist Tanzeum and CARMELINA for Lilly/BI’s DPP-4 inhibitor Tradjenta, as well as full metabolic results from CAMELLIA for Arena/Eisai’s Belviq. In many ways, these readouts came with their own unique atmospheres, considering that (i) Tanzeum has been taken off the market by GSK and its commercial status is unclear, and (ii) non-inferiority vs. placebo on three-point MACE was very much expected with Tradjenta, given that every other DPP-4 inhibitor CVOT has been neutral. To be sure, however, each readout offered immensely important insight on each therapy and its class, also spurring important questions CVOTs in general. By trial:

• In HARMONY, Tanzeum demonstrated a significant 22% risk reduction on three-point MACE vs. placebo (HR=0.78, 95% CI: 0.68-0.90) over a median period of just 1.6 years and in a fully secondary prevention population. Notably, a mean treatment difference in A1c of only 0.52%, favoring Tanzeum, was found at 16 months (p<0.001); in the same timeframe, difference on weight loss was only 0.83 kg (p<0.001).

  • Beyond bolstering the notion that cardioprotection in a class effect of GLP-1 agonists, one of the most important implications of HARMONY results (espoused by many KOLs) is that the mechanism of cardioprotection with GLP-1 agonists is very likely independent of metabolic effects. Given Tanzeum’s mild impact on A1c and body weight, it seems likely that cardioprotection was conferred through an effect of GLP-1 agonism aside from glucose lowering and weight loss. While the trial’s generalizability is limited by its 100% secondary prevention population, it is nevertheless impressive to see such impressive risk reduction in only 1.6 years. To this end, many believed it a shame that GSK withdrew the drug for marketing reasons. Many questions were raised on the trial, its implications, and its design:

    • Will another company acquire and market Tanzeum? Many noted that, as an albiglutide-bound molecule, Tanzeum is more expensive to manufacture than other GLP-1s. Is there any potential to bring Tanzeum back at a lower list price?

    • Which population is Tanzeum best suited for? Given lower efficacy on body weight and glucose, when might it be preferable to use Tanzeum over other GLP-1s? How would this affect pricing across the class?

    • Would GSK have kept Tanzeum on the market if management had known the results?

    • How do these results change cardiologists’ (and endocrinologists’) views on GLP-1 agonists? Can they promote greater uptake of this class for cardioprotection?

    • What do these results tell us about CVOT design? On this note, Dr. John Buse stated: “If you want to win a cardiovascular outcome trial, start with people who have clinical cardiovascular disease and an A1c above 8%.” However, limiting enrollment to those with baseline CVD limits the applicability of those results in the real world. How will the field weight these competing interests going forward?

• In CARMELINA, Tradjenta demonstrating resounding neutrality vs. placebo on the primary endpoint of three-point MACE (HR=1.02, 95% CI: 0.89-1.17, p=0.0002 for non-inferiority, p=0.7398 for superiority). Perhaps more notably, hospitalization for heart failure trended solidly in favor of Tradjenta (HR=0.90, 95% CI: 0.74-1.08, p=0.2635 for superiority), a key result given lingering concern surrounding the DPP-4 class and HHF risk. This was an extremely reassuring result to see. Also of note, a significant 14% reduction in time to first occurrence of albuminuria progression was found (HR=0.86, 95% CI: 0.78-0.95, p=0.0034), though commentator Prof. Philip Home cautioned against extending this finding to a “microvascular benefit” with linagliptin, given that albuminuria is not a market of microvascular complications.

  • The most important class implication of these findings, in our opinion, was the reassuring safety on hospitalization for heart failure. As a reminder, a concerning safety signal on this endpoint was found with AZ’s Onglyza (saxagliptin) in the SAVOR-TIMI trial. In his summary of the trial, Dr. Bernard Zinman also asserted that CARMELINA fills an important gap in data for the DPP-4 inhibitor class, as 63% of the enrolled participants had kidney disease (the next closest is 29% in the EXAMINE trial for Takeda’s Nesina). Here were some of our questions

    • Given these results were predicted by many, what is the value of CVOTs for DPP-4s and established diabetes classes in general? Will the heart failure and albuminuria findings impact clinical practice?

    • How will these results be compared and contrasted with those from the CAROLINA CVOT, which puts Tradjenta up against SU glimepiride? Topline findings from that study are expecting in 2019 (likely early in the year).

• Lastly, in CAMELLIA, Belviq (lorcaserin) was neutral vs. placebo on three-point MACE (HR=0.99, 95% CI: 0.85-1.14), as presented at ESC 2018, but achieved superiority on the microvascular endpoint of persistent microalbuminuria, diabetic retinopathy, and diabetic neuropathy in patients with diabetes at baseline (HR=0.79, 95% CI:0.69-0.92, p=0.001). Moreover, an interesting sub-analysis suggested that patients taking lorcaserin were significantly less likely to start a new glucose-lowering agent (GLA) in the first year of the trial compared to those taking placebo (RR=0.65, 95% CI: 0.58-0.73, p<0.0001) and significantly more likely to stop taking a GLA during this same timeframe (HR=1.28, 95% CI: 1.08-1.51, p=0.005). All of this comes in the context of Belviq’s significant reduction on incident type 2 diabetes in those with prediabetes at baseline (HR=0.81, 95% CI: 0.66-0.99, p=0.038), which was the primary glycemic outcome from the study.

  • Despite positive microvascular and glycemia results, this data further demonstrates the enormous unmet need for better obesity pharmacotherapy. Indeed, independent commenter Dr. Naveed Sattar expressed this view during his presentation: “We still need safe weight loss drugs…We have excellent cholesterol, blood pressure, and diabetes drugs. Things are really, really good there. Yet, obesity and associated complications are rising, in part due to the obesogenic environment…Obesity is now, as I see it, the biggest challenge to those of us in the metabolic profession.” The fact remains that weight loss in CAMELLIA was low at study end, at ~4.2 lbs with Belviq vs. placebo, indicating limited overall efficacy over ~three years. We would love to see an efficacious obesity drug with demonstrated CV benefit, especially given the safety issues surrounding the first generation of obesity therapies. To be sure, we are optimistic about the future, with Novo Nordisk’s highly-anticipated SELECT CVOT for semaglutide 2.4 mg in obesity now recruiting. As we see it, CAMELLIA in no way bodes poorly for the massive (n=~17,000) obesity trial both because (i) the GLP-1 agonist is considered a far more efficacious weight loss agent and (ii) as a GLP-1, semaglutide carries the promise of cardioprotection via non-glycemic mechanisms. We eagerly await results following September 2023 completion.

6. GLP-1 CVOT Post-Hocs (SUSTAIN-6, ELIXA, EXSCEL) Reignite Conversation on Possibility of Renal-Protection and Bolster the Case for Cardioprotective Class Effect

• Post hocs from SUSTAIN 6 and ELIXA, the CVOTs for semaglutide and lixisenatide, re-ignited the conversation around renoprotection as a potential effect of GLP-1 agonists, particularly in patients with worsening renal function. First, a SUSTAIN 6 analysis (for Novo Nordisk’s Ozempic) demonstrated a decrease in albuminuria (specifically, UACR) with semaglutide vs. placebo, regardless of baseline renal function. Interestingly, a dose-dependent effect was found on the composite renal outcome, as 1.0 mg of semaglutide gave a significant 48% reduction in new or worsening nephropathy vs. placebo (HR=0.52, 95% CI: 0.33-0.80, p<0.01) while the improvement conferred by the 0.5 mg dose was not significant but still trended strongly in favor of semaglutide (HR=0.76, 95% CI: 0.52-1.11). In ELIXA (for Sanofi’s Adlyxin), participants with microalbuminuria who were randomized to lixisenatide saw UACR fall an average of 6.4% from baseline over the course of the 108-week trial, vs. a 14.8% increase in UACR for those on placebo (p=0.05). Similarly, in participants with macroalbuminuria at baseline, UACR fell an impressive 31.6% with lixisenatide, compared to an 11% rise in UACR with placebo (p=0.0084) – potentially pointing to stronger renal protection in those with more advanced nephropathy. However, these results are far from conclusive, especially considering that neither study demonstrated a significant improvement in eGFR (generally considered the more meaningful indicator of renal function/microvascular complications). In fact, 0.5 mg semaglutide was associated with a numerically larger drop in eGFR vs. placebo in those with normal renal function by the end of the 104-week trial (-9.6 vs. -7.4 ml/min/1.73 m²), as was 1.0 mg semaglutide (-8.6 vs. -6.5 ml/min/1.73 m²); this group – those with normal renal function – saw the largest decreases in eGFR. This finding is interesting in that it is similar results from the ELIXA post hoc. Altogether, our sense is that GLP-1 agonists have offered less consistently compelling evidence for renal-protection than SGLT-2 inhibitors, but we certainly believe this area deserves further investigation – particularly given that treatment for diabetic nephropathy remains an area of immense unmet need. Moreover, options for glucose lowering therapy in those with impaired renal function remain limited. If GLP-1s can offer both powerful glucose-lowering and renal protection to people with impaired kidney function, it would be a huge win for patients and providers alike. 

• In yet another vote of confidence for cardioprotection as a class effect of GLP-1 agonists, a highly-anticipated EXSCEL post hoc (for AZ’s Bydureon) adjusting for open-label drop-in diabetes medication use shifted the trial’s narrow miss on CV superiority (HR=0.91, 95% CI: 0.83-1.00) to nominal superiority (HR=0.87, 95% CI: 0.78-0.97, p=0.011). Over the past year, we’ve heard dozens of thought leaders assert that EXSCEL does far more to support cardioprotection as a class effect of GLP-1s than refute it – and also that EXSCEL in and of itself is a very positive trial. By and large, the trial’s statistical miss has been attributed to EXSCEL’s pragmatic design, enrollment of a larger primary prevention population, use of complicated reconstitution kits, and lack of a run-in period to exclude those with low adherence (even just this last one could make results quite different from other trials that did exclude low adherence). Perhaps more than anything, this post-hoc raises broad questions about CVOTs and clinical trial design. Make sure to check out our full report from the recent FDA Advisory Committee on the 2008 CVOT mandate for a deeper dive into many of these:

  • How should manufacturers balance pragmatism with the intensive care in clinical trials? Surely, real-world data and pragmatic trials offer a more generalizable idea of the effect of therapies in normal clinical practice. Yet, as demonstrated by EXSCEL, taking this approach may make it far more difficult for manufacturers to demonstrate superiority.

  • What comparative value do CVOTs have? Given significant variation in design and enrolled population, our sense is little to none. Going forward, how can trial design be standardized to better inform clinical decision making?

  • Should different statistical methods and/or post-hocs be considered by regulatory agencies in label updates? Perhaps this could encourage pragmatic trials that are more useful to patients.

  • How do patients view clinical trial and post-hoc data? How can results from these trials become more accessible to patients and even HCPs?

Diabetes Technology

1. Smart Pens: With Novo Nordisk 2019 Launch Plans, Connected Insulin Dose Capture Landscape Is Gaining Commercial Steam

• Novo Nordisk made waves on day #1 with the announcement that it will launch smart reusable pens and dose capture devices in 2019. Already-CE-marked, NFC-enabled, reusable smart pens NovoPen 6 and Echo Plus (details here) are set to launch in key markets in 1Q19, followed by a reusable, Bluetooth pen attachment (cap, clip, sleeve?) for disposable, pre-filled FlexTouch pens later in 2019. NovoPen 6 uploading to the Glooko Transmitter via NFC was demoed in a remote corner of the Novo Nordisk booth. Novo Nordisk has also established data partnerships with Dexcom, Glooko, and Roche, along with the promise that more would follow – each of these partners will integrate Novo Nordisk insulin injection data into their platforms, and in some cases, build out decision support based on it. For example, we picked up in the exhibit hall that mySugr plans to pull injection history into its app via NFC, and Glooko told us about plans to integrate insulin data “broadly,” allowing patients to upload insulin data into the mobile app (though noting that NFC features on phones can be “cumbersome”). This also lays important groundwork for HCP- and patient-facing decision support, via all the partnerships and especially with Dexcom’s acquisition of TypeZero and subsequent 3Q18 commentary. Novo Nordisk’s partnership approach resembles that of Lilly, who is also partnered with Dexcom and plans to launch its own smart pens and insulin titration in late 2019/early 2020. Sanofi is working with Common Sensing’s Gocap in a pilot, but has not officially committed to a commercial launch like Novo Nordisk or Lilly. Pockets of patients and providers have had access to devices such as Companion Medical’s first-to-market InPen, Common Sensing’s GoCap, and Pendiq’s smart pen, and depending on business models and cost, millions of people worldwide could eventually own connected pens or pen attachments. We’re eager to see this happen and particularly what it can do for provider-patient relationships, clinical and patient-facing decision support, and above all, patient outcomes. See our smart pen/cap competitive landscape here.

• Not only are more people poised to benefit from connectivity, but data management platforms are beginning to integrate insulin injection data into their reports. We saw sample displays with CGM and insulin data in Roche’s Smart Pix software and Diasend – it instantly became significantly easier to diagnose areas for improvement (e.g., take a bolus sooner before a meal).

2. CGM Updates: FreeStyle Libre 2 Debuts, with Added Alarms; G6 Launches in Europe; Medtronic’s Enhanced Pattern Snapshot

• The major CGM news of EASD came on Day #1, when Abbott announced a CE Mark for FreeStyle Libre 2. The second-gen FreeStyle Libre adds Bluetooth and optional, customizable hypoglycemia (60-100 mg/dl) and hyperglycemia (120-400 mg/dl) threshold alarms. The new blue reader alarms users with sound or vibration alerts, prompting a manual scan to obtain the real-time reading – i.e., FreeStyle Libre 2 uses Bluetooth to ping the reader with an alarm, but does not send the real-time glucose/trend continuously. Abbott’s exhibit hall booth was packed with attendees eager to see the device in action. The on-body size and disposable form factor are identical to the first-generation. Remarkably, sensor/reader pricing is the same for FreeStyle Libre 2 – indicative of a huge access investment from Abbott, since adding Bluetooth certainly adds cost. Like gen one in Europe, FreeStyle Libre 2 is approved for 14-day wear, one-hour warmup, and non-adjunctive use without fingersticks in patients 4+ years. Accuracy is slightly improved over the current EU version, with a 9.5% MARD vs. YSI. FreeStyle Libre 2 launched in Germany in tandem with EASD, to be followed by a gradual European rollout. US timing is unclear. We’re glad to see the addition of alarms, which should enhance safety in hypoglycemia-prone users and potentially boost real-world time-in-range in many users with frequent hyperglycemia (i.e., alarms for hyperglycemia, especially post-meal, will flag an intervention opportunity immediately, rather than a user seeing it whenever the next manual scan occurs). FreeStyle Libre 2 is a key competitive move and a stepping stone to a future goal of continuous real-time glucose communication via Bluetooth (i.e., alarms + real-time glucose value without needing to scan).

• In its first international conference following June’s CE Mark, Dexcom announced the German launch of G6. At the time of the session, this complemented the no-calibration sensor’s availability in the US, UK, Austria, and Switzerland. (Per 3Q18, G6 is now available in 14 countries.) Dexcom also shared that the Clarity mobile app would launch internationally for the first time in the “coming weeks,” an excellent win for OUS users – to date, they have only had access to Clarity on the web. Notable new data from the symposium examined real-world G6 data from the US launch, spanning ~25,000 patient days (May 1-July 31). Those switching from G5 to G6 have seen: (i) an 18% reduction in time <70 mg/dl, from 3.4% to 2.8% (-9 mins/day); and (ii) a 36% reduction in time <55 mg/dl, from 1.1% to 0.7% (-6 minutes per day). Though these are small absolute changes, they still imply G6 is moving the needle at the extremely low end – even in people previously using G5. Those with the urgent low soon feature turned ON have seen a notable 36% reduction in time <70 (2.8% vs. 4.4%; -23 minutes/day).

• On the software front, Medtronic presented Pattern Snapshot 2, an improved clinical decision support tool for professional CGM. Projected to launch to a small number of clinics in 4Q18, the second version of this software: (i) gives a PCP up to four detected patterns (filtered by possible cause based on medication information, now including oral drugs); (ii) provides tailored recommendations to address the patterns; and (iii) moves the glucose trace higher up on the page for easier glance-ability. For example, when a patient on metformin and glimepiride shows a pattern of hypoglycemia, the PCP is advised to counsel the patient on possible causes (“less food intake? More exercise? Alcohol consumed?”) and also consider adjusting medications (“Reduce or stop SU; Consider replacing SU with a non-hypoglycemia inducing medication”). What a tool for PCPs!

3. CGM Data from People Without Diabetes Shows What “Normal” Is: Mean Glucose of 99 mg/dl, CV of 17%, 97% Time-in-Range (70-140 mg/dl)

• Jaeb’s Dr. Roy Beck presented on behalf of Dr. Anne Peters: 10 days of blinded Dexcom G6 data were analyzed from 153 children (6+ years) and adults without diabetes or prediabetes. Across the entire cohort, mean glucose was 99 mg/dl, and mean standard deviation (SD) was 17 mg/dl, resulting in a coefficient of variation (CV; SD divided by mean) of just 17%. Time-in-tight-range of 70-140 mg/dl was 97% (!) in this group of people without diabetes (2% were >140 mg/dl, 1% were <70 mg/dl). These data really show how far we have yet to go in normalizing blood glucose in diabetes! Put differently, many would put benchmarks for excellent management in type 1 diabetes at a mean glucose of <154 mg/dl, time-in-range of 70%+ (with the wider 70-180 mg/dl), and a CV of <33% - this suggests that even automated insulin delivery is still pretty far from what people without diabetes experience. At the same time, data from a Dr. Rich Bergenstal poster here demonstrated that a time-in-range of 87% is attainable with mean glucose <166 mg/dl and CV <33%!

• Other interesting findings from the study, from Dr. Beck:

  • Mean glucose was 104 mg/dl in individuals over the age of 60 years – 5 mg/dl higher than the group mean. This could indicate age-related glucose intolerance, or perhaps the 60+ group had higher A1cs than did the other age groups in the study by chance.

  • Over one-third of the participants spent ≥30 minutes below 70 mg/dl each day.

  • Meal and activity data had the same takeaway: crazy low variability.

4. Strong AID Data Continues: Cambridge’s HCL 12-Week RCT, Cambridge’s Fully Closed Loop in Inpatient Type 2s, UVA 3-day Ski Camp Study in 6-17-Year-Olds on Control-IQ

• For a conference typically is focused more on diabetes therapy, there was a handful of new data on automated insulin delivery systems. We saw extremely positive A1c and time-in-range results from the Cambridge group’s 12-week, randomized trial comparing hybrid closed loop (n=46) to sensor-augmented pump (n=40) in patients with a high baseline A1c of 8.2%-8.3%. Notably, the robust, randomized home study included adults, adolescents, and children (6+ years) and represents the largest randomized study of closed-loop use in outpatient settings so far. Hybrid closed loop drove a +3 hour/day time-in-range gain and had a 0.4% A1c advantage vs. SAP. Results were simultaneously published in The Lancet. It was also only the third randomized study to report A1c outcomes. In addition to this strong data, the Cambridge group presented a poster of a randomized parallel study comparing its fully automated closed loop system (n=10) vs. conventional insulin therapy (n=9) in type 2 adults undergoing hemodialysis while in the hospital. Patients on closed loop spent a remarkable ~9.3 more hours in-range than the control group, further supporting the very strong inpatient closed loop data we saw at ADA. The powerful data speaks volumes about how much value AID will bring to the hospital, as well as how poor hospital care is right now (see Adam’s recent diaTribe piece). Another highly-anticipated poster came from UVA team/Stanford/BDC, which depicted very strong results from a three-day ski camp study investigating the Tandem t:slim X2 with Control-IQ hybrid closed loop in adolescents (n=24; 13-17-year-olds) and children (n=24; 6-12-year-olds). We were particularly struck by the interaction between age and system performance – adolescents saw far greater reductions in time in hyperglycemia across all thresholds observed, as well as larger increases in time-in-range and stronger decreases in average sensor glucose. We’ll be very interested to see if these age differences persist the pivotal study (fully enrolled; summer 2019 launch expected, per 3Q18). With the iCGM pathway already in place in the US and substantial buzz surrounding iPump and iAlgorithm/iController pathways, the future of AID systems with interoperable, plug-and-play components is exciting. We expect to be seeing a lot of movement and accelerated innovation here and look forward to watching both of these groups bring their systems to market.

Big Picture

1. Glycemic Variability Tied to Hypo Risk, Regardless of Mean Glucose – Need More Studies to Prove Variability Link to Long-Term Complications; KOLs Continue Push for Greater Emphasis on CGM Metrics

• In the closing moments of EASD, Prof. Thomas Danne called for longitudinal studies to demonstrate the link between glycemic variability and long-term complications. Regarding the correlation between glycemic variability/time-in-range and long-term complications, Prof. Danne outlined a number of cross-sectional and observational studies in his talk: (i)  time-in-range correlates with retinopathy; (ii) glucose standard deviation correlates with alterations in the white matter structure in type 1 children; (iii) MAGE correlates with variability in neointimal thickness and frequency of uncovered stent struts (CV markers); (iv) and higher day-to-day fasting glycemic variability correlated with higher risk of severe hypoglycemia (and all-cause mortality) in DEVOTE. We’d add Dr. Roy Beck et al.’s recent analysis showing that SMBG-derived “time-in-range” correlated with microvascular outcomes in the DCCT, to that list. These studies may lack the rigor to sway regulators and payers on their own, but in aggregate, they begin to paint a convincing picture of the deleterious effects on un-physiological fluctuation of blood glucose in people with diabetes. Still, Prof. Danne claimed there’s a need for longitudinal studies – we see value in retrospective analyses (such as Dr. Beck and colleagues’ from DCCT) alongside prospective studies with biomarkers and validated risk prediction scores to make the case even stronger.

• Noteworthy speakers at EASD including Dr. Bergenstal, Prof. Danne, Dr. Pratik Choudhary, and Prof. Brian Frier lent their voices to the cases for time-in-range and hypoglycemia. Drs. Bergenstal and Danne both suggested that time-in-range and hypoglycemia will be the yardstick of diabetes therapy, with the latter even proposing a nomenclature borrowed from blood pressure: “72/5,” where “72” is the percent time between 70-180 mg/dl, and “5” is the percent time below 70 mg/dl. Time-in-range alone is not sufficient to describe a person’s glycemia, but we like Prof. Danne’s system, especially because the percent time in hyperglycemia (>180 mg/dl) follows naturally with some quick mental math. VU Medical Center’s Prof. Frank Snoek agreed, proposing that hypoglycemia should be monitored as a routine part of consultation. In a Lilly-sponsored symposium, Prof. Choudhary wondered why type 2s are frequently not taught how to adapt insulin to changing health status – a huge hypoglycemia risk – and Prof. Frier called it “madness” to target A1cs of 6%-7% or give long-acting SUs to elderly people. We are pleased that the conversation continues to shift, giving Prof. Danne the confidence to say that, “Clearly, A1c is on the way out.”

• A piece of new data on glycemic variability at EASD came from a Dr. Rich Bergenstal/Novo Nordisk poster pointing to a strong correlation with hypoglycemia. Real-world CGM data uploaded to the Novo Nordisk Cornerstones4Care app from 137 people with diabetes showed that reducing glycemic variability correlates with decreased time in hypoglycemia, regardless of mean glucose. At a high level, patients with a coefficient of variation (CV) ≥37% in the study spent >3x the amount of time <54 mg/dl as those with CV <37% (1.3% vs. 0.3%). Those with CV ≥48% spent 19x more time <54 mg/dl than those with CV ≥37% (6%). This finding is by no means revolutionary (nor from an RCT, which would probably be unethical), but it does lend support to the idea that glycemic variability should be minimized to prevent episodes of hypoglycemia, which have in turn been shown to correlate with severe hypoglycemia events. The authors concluded that there is “likely significant value” in reducing glycemic variability regardless of mean glucose levels so as to decrease the risk of hypoglycemia – hard to argue with that, and we hope data like this continue to shift standard of care.

2. Optimism Surrounding Diabetes Prevention Programs; Engaging Multiple Stakeholders in Prevention, Including Politicians and Communities Involved in Related Disease States

• EASD was filled with fascinating commentary on diabetes prevention. Dr. Jaako Tuomilehto delivered the prestigious Claude Bernard Lecture, providing a tour de force review of over 120+ years of diabetes prevention research. Dr. Tuomilehto’s perspective was generally optimistic, pointing to the efficacy of prevention programs, the proliferation of such programs over recent years, and the declining incidence of new diabetes cases observed in the US and some other countries. Later in the day, Universidad Autónoma de Madrid’s Professor Rafael Gabriel expressed similar enthusiasm for diabetes prevention programs, but took it a step further, asserting that no further primary prevention trials are needed to support the use of DPPs. Prof. Gabriel explained that evidence now unequivocally supports that effective DPPs can result in a 50% decrease in the risk of type 2 diabetes – a degree of efficacy unmatched by any diabetes drug currently on the market. Prof. Gabriel noted eight specific factors that lifestyle interventions should target and drew a distinction between population-level (e.g., soda tax) and high-risk (e.g., DPP) strategies. Ultimately, Prof. Gabriel argued that calls-to-action for diabetes prevention from non-governmental organizations are not sufficient – multiple stakeholders, including politicians need to be involved. To this end, EASD President Dr. Juleen Zierath officially launched the European Diabetes Forum during her presidential address to commence EASD 2018. The Forum aims to catalyze the effective translation of scientific research into policy actions that can be implemented broadly on a national level, engaging “all stakeholders in the diabetes landscape.” We wonder if the Forum should stop at just diabetes stakeholders – in Dr. Tuomilehto’s second lecture of the day, he forecasted that cancer will soon have a similar standing as CV disease as a full-fledged and incredibly urgent complication of diabetes. Moreover, Dr. Tuomilehto pointed to Finnish registry data showing a strong link between Alzheimer’s disease and diabetes. Perhaps the combined efforts of these communities could grab more varied stakeholders’ attentions.

GLP-1 Agonists and DPP-4 Inhibitors

Symposium: HARMONY-Outcomes

Introduction and Background

Adrian Hernandez, MD (Duke Clinical Research Institute, Durham, NC, USA)

Dr. Adrian Hernandez opened this highly-anticipated session with an introduction to and background on the HARMONY-Outcomes trial for GLP-1 agonist albiglutide. The trial began in July 2015 and concluded in March 2018; notably, GSK withdrew support for the commercial product Tanzeum in July 2017 but allowed and provided for the continuation of the trial. Dr. Hernandez first provided an overview of the design of previous CVOTs for the GLP-1 class in order to properly place HARMONY within this broader context, noting some key design differences that differentiated each of the previous CVOTs: ELIXA (for lixisenatide) exclusively enrolled patients with acute coronary syndrome, while EXSCEL (exenatide) also enrolled primary prevention patients in addition to a majority population with established CVD. Meanwhile, LEADER (liraglutide) boasts the longest median duration of follow-up of any of the studies (3.8 years). HARMONY, as we note below, exclusively enrolled patients with type 2 diabetes and established CVD. We continue to be struck by the heterogeneity of CVOT design; because of the inevitable comparisons that are drawn between their results, we hope for a significantly greater push towards standardization of these trials. Dr. Hernandez also offered context on the properties of albiglutide, which has a half-life of approximately 5 days, allowing for effective once-weekly dosing of the agent. Molecularly, albiglutide shares 97% homology with human GLP-1, placing it in a category with semaglutide and liraglutide (lixisenatide and exenatide are based off of the exogenous exendin-4). More specifically, albiglutide is composed of two linked copies of recombinant human GLP-1 fused to recombinant human albumin, which stabilizes the complex and extends the half-life of the therapy (as detailed below, it also makes albiglutide more expensive to manufacture than other GLP-1s).

Design, Methods, and Baseline Characteristics

Dr. Jennifer Green, MD (Duke University, Durham, NC)

Dr. Jennifer Green reviewed the design, methods, and baseline characteristics of HARMONY. A total of 9,643 (21% North American, 18% Latin American, 22% Eastern European, 36% Western European and South African, and 4.0% Asian Pacific) participants were randomized 1:1 to albiglutide (starting at 30 mg, with 50 mg available after week 5) or placebo. Participants visited an HCP every four months; A1c was measured each time while liver function and eGFR were tested every other visit. Standard of care, as defined by local practice and published guidelines, was continued for all participants throughout the study. Eligibility requirements included: (i) type 2 diabetes; (ii) age ≥40 years old; (iii) A1c >7.0%; and (iv) established atherosclerotic cardiovascular disease (coronary artery disease, cerebrovascular disease, or peripheral arterial disease). 1,330 of 10,793 screened participants either left the trial prior to randomization or were excluded due to poor renal function (eGFR <30 mL/min/1.73 m² or need for renal replacement therapy) or other exclusion criteria. At baseline, participants were 64 years old on average, ~70% male, and had an A1c of 8.7%-8.8%. The patient population had significant CV disease history at baseline: ~70% had coronary artery disease, 86% had hypertension, and 20% had heart failure. Because the population was relatively sick, medication use at baseline was high: approximately 84% were on a statin, 50% an ACE inhibitor, 33% an ARB, 67% a beta blocker, 77% aspirin, 74% metformin, 29% an SU, 16% a DPP-4, and 7% an SGLT-2 inhibitor. Median follow-up on the primary endpoint was 1.6 years.

• In terms of outcomes, the primary endpoint was the now-standard, adjudicated three-point MACE (non-fatal MI, non-fatal stroke, or CV death). Secondary vascular outcomes were the individual components of MACE, MACE or urgent revascularization for unstable angina, CV death or hospitalization for heart failure, all-cause mortality, and a composite of microvascular events including need for renal transplant or dialysis, new diabetes-related blindness, laser photocoagulation, anti-VEGF treatment or vitrectomy. Secondary metabolic outcomes were change in A1c, weight, time to initiation of insulin and prandial insulin, and proportion of subjects achieving all of A1c ≤7% with no severe hypoglycemia incidents and weight gain <5% of body weight. Safety outcomes were all-cause death, change in eGFR, severe adverse events, adverse events of pre-specified special interest, and change in blood pressure and heart rate.

Effects on Glycaemic and Renal Outcomes

Stefano del Prato, MD (University of Pisa, Italy)

Dr. Stefano del Prato presented glycemic and renal outcomes, demonstrating that albiglutide had a modest effect on A1c and body weight over the course of HARMONY. Baseline A1c was 8.8% in the albiglutide group and 8.7% in the placebo group; at eight months, the placebo-adjusted treatment difference with albiglutide was 0.63% (95% CI: -0.69%, -0.58%, p<0.001). By 16 months, this had attenuated to a 0.52% treatment difference (p<0.001). At study end, 28% of the albiglutide group had an A1c ≤7.0%, compared to 17% in the placebo group. Turning to body weight, placebo-adjusted treatment difference with albiglutide was only 0.66 kg (95% CI: -0.83, -0.49; p<0.001) at eight months, and rose to 0.83 kg (95% CI: -1.06, -0.60; p<0.001) at 16 months. Dr. del Prato characterized this difference as “moderate though significant,” and, overall, these results support that cardioprotection with albiglutide was likely not conferred through metabolic effects. At last dispensing, 50% of the albiglutide group had been escalated to the higher, 50 mg dose (compared to 63% on placebo); premature discontinuation of study drug occurred in 24% of the albiglutide group vs. 27% of the placebo group. Both groups saw an increase in concomitant medication use, including DPP-4 inhibitors and SGLT-2 inhibitors; precise numbers were not given, but it was clear that, in particular, more SGLT-2 inhibitors, SUs, and insulin were added to the placebo group. In the albiglutide group, time to initiation of insulin was decreased by 58% vs. the placebo group (HR=0.42, 95% CI: 0.33-0.53, p<0.001 for superiority); time to initiation of prandial insulin in those on basal insulin at baseline was reduced by 29% (HR=0.71, 95% CI: 0.51-0.99, p=0.043). However, Prof. David Matthews later commented that this is to be expected in such a trial. Very encouragingly, despite low overall incidence of severe hypoglycemia events, there was a significantly lower incidence in the albiglutide group (0.7%) vs. the placebo group (1.2%), giving a risk ratio of 0.56 (95% CI: 0.36-0.87).

• In the albiglutide group, 26% of participants achieved a final metabolic composite of (i) final visit A1c ≤7.0%; (ii) no severe hypoglycemia episodes; and (iii) weight gain <5%, compared to 15% of the placebo group (p<0.001 for comparison). Dr. del Prato characterized this as “quite good,” though absent the CV benefit, we think the components of this endpoint could be more compelling (e.g. weight loss rather than mild weight gain). Moreover, we’d be very interested to see how this compared to results from other GLP-1 agonist CVOTs over a similar timeframe. 

• Turning to microvascular endpoints, Dr. del Prato honed in on a significant decrease in eye events (41 with albiglutide vs. 60 with placebo, in 33 and 52 patients, respectively). This included reductions in new diabetes-related blindness (4 vs. 14 events), anti-VEGF treatment (16 vs. 26 events), photocoagulation (24 vs. 31 events), and vitrectomy (6 vs. 13 events). Little granularity was offered on renal outcomes; there was only a hemodialysis endpoint displayed, which showed four events with albiglutide vs. two with placebo. However, we do know that there was no impact on eGFR, which isn’t surprising for a 1.6 year median follow-up. At 16 months, there was only a 0.4 ml/min/1.73 m² difference in eGFR (narrowly favoring placebo; p=0.315)

Effects on Cardiovascular Outcomes

John McMurray, MD (University of Glasgow, UK)

Dr. John McMurray took to the stage to deliver the highly-anticipated cardiovascular outcomes of HARMONY. On the primary endpoint of three-point MACE (non-fatal MI, non-fatal stroke, CV death), albiglutide delivered a 22% relative risk reduction vs. placebo (HR=0.78, 95% CI: 0.68-0.90, p<0.0001 for non-inferiority, p=0.0006 for superiority) with 428 patients (9%) experiencing an event in the placebo group vs. 338 (7%) in the albiglutide group – wow! The excitement (and, we think, surprise) in the room was palpable, and only enhanced by the fact that median follow-up was a relatively brief 1.6 years. Surely, this result is at least partially attributable to HARMONY’s relatively “sick” study population of 100% participants with baseline CVD (the event rates were a quite high 5.87 and 4.57 per 100 person-years in the placebo and albiglutide groups, respectively). Nevertheless, HARMONY offers compelling evidence of cardioprotection as a class-effect of GLP-1s broadly and albiglutide specifically. Moreover, the significance of the result held up to all sensitivity analyses, including one accounting for deaths of unknown causes and one assessing publicly available data for patients who withdrew consent for study follow-up. More specifically, the reduction in MACE was largely driven by a significant 25% reduction on non-fatal MI (HR=0.75, 95% CI:0.61-0.90), with 240 events in the placebo group (3.26 per 100 patient-years) vs. 181 events in the albiglutide group (2.43 per 100 patient-years). Results for the other independent components of MACE did not reach statistical significance but trended toward benefit for albiglutide: HR=0.93 (95% CI: 0.71-1.19) for CV death and HR=0.86 (95% CI: 0.66-1.14) for non-fatal stroke. Of note, the authors of HARMONY’s publication postulated that a longer study duration would have demonstrated a significant reduction in CV death with albiglutide vs. placebo.

• On the secondary composite endpoint of CV death or hospitalization for heart failure, results heavily favored Tanzeum but did not meet significance (HR=0.85, 95% CI: 0.70-1.04). That said, it should be noted that results shifted further in favor of Tanzeum when heart failure was lumped in with CV death compared to CV death alone, with Dr. McMurray noting that this shift was due to “significantly fewer hospitalizations for HF with albiglutide” – perhaps the opposite effect to be expected with a GLP-1 agonist.

• To close out the cardiovascular outcomes, Dr. McMurray shared a sobering statistic, putting these “positive” results into perspective: The number needed to treat with albiglutide over a median of 1.6 years to prevent one MACE event is 50. Of course, preventing one MACE out of 50 in just a year and a half is an improvement, but, by the same logic, 98% of MACE events during that time frame will still occur. Surely, further improvement here is a goal we can all get behind.

Effects on Safety Outcomes

Chris Granger, MD (Duke University, Durham, NC)

Duke’s Dr. Chris Granger presented safety and adverse event data from HARMONY. Prespecified adverse event analysis for HARMONY focused on two categories of adverse events that were of special interest: those arising from the GLP-1 agonist profile (including thyroid and pancreatic cancer, severe hypoglycemia, and serious GI events), and those based off of phase 2/3 studies and regulatory request (such as worsening renal function and retinopathy). Dr. Granger presented data demonstrating that incidence of serious adverse events was balanced between treatment and placebo groups (19.8% vs. 21.7%, respectively) and that no individual serious adverse event occurred in more than 2% of patients. The only adverse event significantly more common in the treatment group was injection site reactions (86 vs. 29 events; HR=2.96, 95% CI: 1.95-4.51). Severe hypoglycemia was significantly less common in the albiglutide group (31 vs. 55 events; HR=0.56, 95% CI: 0.36-0.87) and likely attributable to increased use of insulin in the placebo arm in order to achieve glycemic equipoise. Importantly, no negative signal was detected for retinopathy in HARMONY; in fact, the data trended in favor of albiglutide treatment (HR=0.88; 95% CI: 0.65-1.18). After SUSTAIN-6 demonstrated an association between semaglutide and retinopathy, some HCPs have expressed worry over the potential connection between GLP-1s and this adverse event. While most thought leaders have coalesced around the “early worsening phenomenon” as an explanation of this effect (attributing the retinopathy to rapid improvements in glucose control with semaglutide), HARMONY should nevertheless help assuage these fears.

Summary and Conclusions/Implications

Lawrence Leiter, MD (University of Toronto, Canada)

In offering conclusions and context, Toronto’s Dr. Lawrence Leiter was highly positive about HARMONY’s MACE and MI outcomes and emphasized that all components of the primary endpoint trended in the right direction and were consistent across subgroups. While, he says, the body of data from GLP-1 agonist CVOTs trends strongly in favor of cardioprotection for these agents as a class, there are important differences between both molecules and trial designs that might well be contributing to heterogeneity in results across these trials (see below). Indeed, across the five reported GLP-1 agonist CVOTs, there is statistical heterogeneity on three-point MACE results (I²=59%; p=0.044) despite an overall positive effect (HR=0.88, 95% CI: 0.80-0.96). We note that, in many key aspects, HARMONY and albiglutide are very similar to LEADER and liraglutide, as well as SUSTAIN 6 and semaglutide. For instance, all three agents are based on human GLP-1 and have a longer duration of action, and all three CVOTs enrolled a more sizable (or entirely) secondary prevention population with a higher A1c. Dr. Leiter also emphasized that in HARMONY investigators were free to use any other glucose-lowering therapies (including DPP-4 and SGLT-2 inhibitors); he also highlighted albiglutide’s high homology to human GLP-1 and the fact that this result was achieved despite having relatively very low glycemic potency. Two key limitations, however, were that (i) the relatively short 1.6 year follow-up duration likely limited the trial’s ability to adequately explore mortality and microvascular endpoints and (ii) lipids and urinary albumin were not measured, limiting potential insights into albiglutide’s mechanism of action. We’d add another aspect that Dr. Leiter has previously discussed: The entire population enrolled already had CV disease, limiting the trial’s potential to speak to any potential cardioprotection in lower-risk patients. Dr. David Matthews also addressed this issue, below. Ultimately, Dr. Leiter concluded that HARMONY offers evidence in support of greater utilization of certain GLP-1 agonists to reduce CV risk in patients with type 2 diabetes – we could hardly agree more.

Commentator

David Matthews, MD (University of Oxford, UK)

Professor David Matthews’ largely positive commentary on HARMONY began with brief praise of the study’s design, analysis, and results, with a significant caveat – the study population was “really, really sick.” As a result, he warned against extrapolating these positive results beyond those with established cardiovascular disease and heavy concomitant medication use, noting that the short timeframe for the study should be interpreted with the study population in mind: “With a trial desperate to finish in 1.6 years, indeed you’ll select participants with poor glycemic control and established cardiovascular disease.” To be sure, the generalizability of CVOTs has emerged as one of the biggest, if not the biggest, issues in the realm of diabetes drugs today. As such, Prof. Matthews cheekily complimented those who designed HARMONY for “gaming” well by recruiting patients who were very near the edge of having a MACE event but just far enough away to allow for albiglutide to have an effect. He also translated the study’s statistical language into layman’s terms. As he explained, “If non-inferiority was established, effect on adjudicated MACE would be tested for superiority in a closed testing procedure” simply means analyzing the data using a one-sided (superiority) vs. two-sided (non-inferior) statistical test. Continuing his statistical tangent, Prof. Matthews critiqued the ad-hoc fashion by which p-values are determined for secondary outcomes as “ludicrous.” As it stands, CVOTs must pre-specify their endpoints, including all secondary endpoints, in a hierarchy for analysis. As soon as one fails to demonstrate superiority, all subsequent endpoints can only be evaluated for nominal p-values. As it pertains to HARMONY, the study designers defined CV death as the first secondary endpoint. Seeing as this endpoint only met non-inferiority, the second secondary endpoints of fatal or non-fatal MI could only be evaluated for a nominal, non-inferiority p-value (0.003) despite the fact that it drove the effect on three-point MACE. Prof. Matthews asserted, “We shouldn’t have trials that are skewed by what the investigators think is going to happen…The monotonic nature of the divergence of the Kaplan-Meier curve for MI shows that this is really likely to be a true finding…You’d have to be living on another planet if you didn’t think that there was some real evidential basis for protection against MI with an HR of 0.75.”

• While cardioprotection may be a class effect of GLP-1 agonists, there’s no doubt that heterogeneity exists across other outcomes. Prof. Matthews quantified the sequential p-values associated with weight change as “the square root of no difference…Whatever they’re saying about these things, the fact is that weight did not change with albiglutide.” To further illustrate this point, he pointed to data for Novo Nordisk’s oral semaglutide demonstrating ~6 lbs greater weight loss compared to placebo at the highest dose (14 mg), and greater A1c reduction to boot, arguing that this therapy is clearly superior to albiglutide on these metabolic outcomes. Based on their incredibly different effects on weight and A1c, Prof. Matthews postulated a different mechanism of action for the two GLP-1 agonists, suggesting that oral semaglutide could be passing through the blood-brain barrier. However, he did point out that there are potential applications for a GLP-1 agonist that confers CV protection without weight loss, particularly in thin people and those with type 1 diabetes.

• To conclude, Prof. Matthews extolled GSK’s “highly ethical behavior” in continuing HARMONY through to the end, even after reprioritizing away from diabetes and discontinuing commercial support for Tanzeum. We hope that GSK can find a partner to bring Tanzeum back to the market. We have heard that, because albiglutide is an albumin-bound molecule, it’s more expensive to manufacture, but we remain hopeful that albiglutide could return as a more affordable GLP-1 agonist.

Thought Leader Opinions on HARMONY

Thought Leader Opinions on HARMONY

We reached out to a number of thought leaders in diabetes and cardiology about the positive results from the HARMONY CVOT for GSK’s discontinued GLP-1 agonist Tanzeum (albiglutide), presented in an oral session yesterday. See below for a series of illuminating thoughts on cardioprotection as a GLP-1 class effect, the impact of trial design and molecular differences, and future prospects for Tanzeum. We salute the KOLs for taking time to share these valuable insights.

• Dr. Julio: Rosenstock (Dallas Diabetes Research Center, TX): “This could be a cheaper weekly GLP-1 RA with a 22% reduction in CV events with a niche used to improve glucose control and make life easier for T2DM people on MDI by taking less insulin shots! Taken together, the CV benefits and the impact on MDI mean we can say that the Phoenix has risen again from the ashes!”

• Dr. Daniel Drucker (Mt. Sinai Hospital, Toronto, Canada): “The data clearly dissociate reduction in glucose and body weight (modest with albiglutide) from reduction in MACE and MI. This is very unexpected and provocative data. These results imply different dose response curves for cardioprotection vs. metabolic benefits and bolster the value of the GLP-1 class beyond glucose control. I suspect GSK would not have made the decision to jettison albiglutide more than one year ago if they had a crystal ball that predicted this remarkable result. What company has ever abandoned a drug that reduces MACE events?”

• Dr. Philip Home (Newcastle University, UK): “This is an interesting result because really no one disputes that glucose-lowering and weight loss are less strong with albiglutide, and here in particular as most of the participants do not appear to have progressed from the 30 mg, less effective dose. But the finding at headline level is better than liraglutide in LEADER and pretty comparable to semaglutide in SUSTAIN-6. When, however, I look at all the findings for MACE and its components from LEADER, SUSTAIN-6, HARMONY, and yes even EXSCEL, I find I cannot tell that they do not come from the same population. That raises the issue of mechanism again.  If the GLP-1s are so different on glucose and on body weight, then why are they indistinguishable on CV outcomes? Of course, we already knew it was not glucose-lowering, because that takes too long to affect CV outcomes, but the implication is that there is a different response curve for the CV protective mechanism than for glucose/weight. But this CV mechanism may exist outside diabetes and in type 1 diabetes, so we now need to investigate how these medications perform in people with CV disease who do not have diabetes, and people with type 1 diabetes who have CV disease. I do not suspect that anyone will be interested in acquiring albiglutide from GSK. The albumin is expensive to produce, and the factory has wound down operations. Therefore, the cost of production is high, but then I suppose current GLP-1 prices allow a lot of room for price reduction.”

• Dr. Naveed Sattar (University of Glasgow, UK): “I was surprised by the outcomes given background context of modest effects on some risk factors but, then again we are all learning that some, perhaps many, surrogate risk marker changes cannot predict hard outcomes and so trials remain critical. My take is that some direct effect of this GLP-1 must affect the atherothrombotic process and therefore reduce risk predominantly of MI.

   In terms of future use, it’s hard to predict but given other drugs in the class perform better in terms of weight and glycaemia, many clinicians might still prefer to recommend those drugs (with proven CVD benefit) over albiglutide. However, in the secondary prevention setting, where cardiologists want to lower subsequent CV risk, and have less concerns about glycaemia reductions per se, it makes this drug with its clear MACE benefit a real potential option especially given its once weekly treatment. But overall interesting result and interesting times for the diabetes field – a general boost to the GLP-1 class.”

• Dr. Sanjay Kaul (Cedars-Sinai Medical Center, Los Angeles, CA): “The impact on clinical outcomes trumps impact on surrogate markers! Yes, the glucose-lowering effect might have hindered its uptake, but it is a shame to withdraw the drug for marketing reasons. The fact this is only the second GLP-1 to yield clear-cut cardioprotective benefits questions the wisdom of that decision! I am sure the patients who signed up for this trial were not made aware during the informed consent process about the possibility of withdrawing the drug for marketing reasons! What are the ethical obligations of the sponsor, the DMSB and the investigators to these patients? It is difficult to point out exactly the impact of trial design, population, duration of action or molecular composition on trial outcomes. The placebo event rate in HARMONY is higher than what was observed in LEADER or SUSTAIN-6 and approaches that of the placebo arm in ELIXA, an acute coronary syndrome trial. This could be related to the fact all patients in HARMONY had prior history of CVD compared with 81% in LEADER and 83% in SUSTAIN-6. So far, the only trials with positive evidence of benefit are those that have utilized human GLP1-RA. Thus, molecular composition could conceivably contribute to the benefits. However, the pragmatic design of EXSCEL might have contributed to the suboptimal treatment adherence leading to the trial barely failing to meet superiority. In addition, the clustering of events early during the ELIXA trial (post-ACS cohort) might have masked the ability of the drug to yield a positive outcome.”

• Dr. John Buse (University of North Carolina, Chapel Hill, NC): “The main thing is that it demonstrates that the GLP-1 cardiovascular effect is probably mediated through the GLP-1 receptor. If you want to win a cardiovascular outcome trial, start with people who have clinical cardiovascular disease and an A1c above 8%.”

• Dr. Darren McGuire (University of Texas Southwestern Medical Center, Dallas, TX): “I’m pleasantly surprised by the significant reduction in MACE; this speaks strongly to class effects, and heterogeneity in outcomes across the class I believe may be more to do with chemical derivation and pharmacodynamics than GLP-1 receptor agonism per se. The observed superiority on MACE despite less potency for glucose lowering and less substantial effects on weight loss go a long way to parse these effects from other probable mediators of atherosclerotic vascular disease. I very much hope another company will pick up this baton; it seems to be a valuable option for clinical care and perhaps may force more competitive pricing across the class. I think the heterogeneity in the estimates of effects across the components of the MACE outcomes between LEADER, SUSTAIN 6 and HARMONY are most likely spurious, with limited power in any of the trials for any single component outcome.”

Symposium: Cardiovascular and Renal Microvascular Outcome Study with Linagliptin in Patients with Type 2 Diabetes Mellitus (CARMELINA)

Study Design and Statistical Plan

Robert Toto, MD (UT Southwestern Medical Center, TX)

Ahead of publication, results from the CARMELINA CVOT for Lilly/BI’s DPP-4 inhibitor Tradjenta (linagliptin), which completed in January 2018, were presented (see the announcement). As expected, results on cardiovascular outcomes were neutral, but there was indication of benefit on albuminuria in this very high-risk population. Now, we’re eager for outcomes from the CAROLINA CVOT, which puts Tradjenta up against SU glimepiride – but first, get a rundown of CARMELINA results, starting with the study design. Dr. Robert Toto kicked off this very well-attended session by emphasizing the scope of CARMELINA (n=6,979) as an international trial that randomized patients from 605 sites across 27 different countries; 3,494 were randomized to 5 mg linagliptin and 3,485 to placebo, on top of standard-of-care. HCPs were encouraged to treat all CV risk factors according to local guidelines. Inclusion criteria included baseline A1c between 6.5% and 10% plus high risk of CV events defined by either (i) albuminuria and macrovascular disease or (ii) impaired kidney function. The primary outcome was three-point MACE (CV death, non-fatal MI, and non-fatal stroke), and the key secondary outcomes was a composite renal endpoint (more below). CARMELINA was powered to determine non-inferiority of linagliptin vs. placebo on the primary endpoint.

Baseline Characteristics and Metabolic Outcomes

Steven Kahn, MB, ChB (University of Washington, Seattle, WA)

Dr. Steven Kahn reviewed baseline characteristics and metabolic outcomes. CARMELINA’s population was ~63% male, with an average baseline A1c of ~8.0% and average duration of type 2 diabetes of ~15 years. Per the inclusion criteria, 57% of participants had established CVD, 74% had prevalent CKD, and 33% had both conditions. Only 20% of participants had normo-albuminuria, while the other 80% had either micro- or macroalbuminuria. Moreover, 62% of patients had an eGFR below 60 ml/min/1.73 m². In terms of metabolic outcomes, the linagliptin arm experienced a significant (and expected) reduction in A1c compared to placebo after 3.5 years (treatment difference = -0.36%, p <0.0001). Participants receiving linagliptin were also less likely to require additional glucose-lowering therapies post-baseline, and they were also less likely to initiate or need a dose increase of insulin (HR=0.72, 95% CI: 0.65-0.81, p <0.0001). Consistent with established profile of DPP-4s, no significant weight loss or blood pressure reductions were seen over time.

Cardiovascular Outcomes

Darren McGuire, MD (UT Southwestern, Dallas, TX)

In presenting cardiovascular outcomes, Dr. Darren McGuire opened by revealing that CV death comprised the majority (52%) of the 854 MACE events, followed by non-fatal MI (34%) and stroke (14%). He emphasized a high placebo event rate of 5.63/100 person-years, nearly three times the CVD equivalent. On three-point MACE, linagliptin was neutral vs. placebo (HR=1.02, 95% CI: 0.89-1.17, p=0.0002 for non-inferiority, p=0.7398 for superiority). This remained true when hospitalization for unstable angina was added (HR=1.00, 95% CI: 0.88-1.13, p<0.0001 for non-inferiority, p=0.9598 for superiority). Sensitivity analyses indicated no significant differences in risk for major CV events, and splitting MACE by select baseline characteristics found no significant effects, though there was a moderate trend with A1c (p value for interaction=0.0407 when splitting baseline A1c at 8%; lower A1c favors linagliptin more strongly). Individual endpoints were also neutral, including (i) CV death (HR=0.96, 95% CI: 0.81-1.14), (ii) non-fatal stroke (HR=0.88, 95% CI: 0.63-1.23), and (iii) non-fatal MI (HR=1.15, 95% CI: 0.91-1.45). All-cause mortality was also neutral (HR=0.98, 95% CI: 0.84-1.13).

• Perhaps the most notable endpoint was hospitalization for heart failure, which trended solidly in favor of linagliptin (HR=0.90, 95% CI: 0.74-1.08, p=0.2635 for superiority). In the placebo group, 226 patients had events vs. 209 patients in the linagliptin group. Given the complex history of DPP-4 inhibitors and heart failure, this was an extremely reassuring and overall excellent result to see.

Renal Outcomes

Vlado Perkovic, MD (Georgia Institute for Global Health, Sydney, Australia)

Dr. Vlado Perkovic presented the renal outcomes from CARMELINA. On the key secondary renal endpoint, a composite of time to first occurrence of: (i) sustained decrease of ≥40% in eGFR from baseline (504 total events), (ii) sustained end-stage kidney disease (127 total events), or (ii) death due to kidney disease (2 total events), the study found non-inferiority with linagliptin vs. placebo (HR=1.04, 95% CI: 0.89-1.22, p=0.6164). 327 events were experienced in the linagliptin group vs. 306 with placebo. After removing sustained decrease in eGFR from this composite (which accounted for a majority of the first events in the key secondary endpoint), the hazard ratio shifted measurably in favor of linagliptin but remained non-significant (HR=0.87, 95% CI: 0.69-1.10, p=0.2371). Not included in either of these endpoints was time to first occurrence of albuminuria progression (change from normo- to micro- or macroalbuminuria, or from micro- to macroalbuminuria), for which linagliptin demonstrated a significant reduction vs. placebo (HR=0.86, 95% CI: 0.78-0.95, p=0.0034) with 819 events in the placebo event and 763 events in the linagliptin group. This endpoint then drove the superiority of linagliptin vs. placebo (HR=0.73, 95% CI: 0.78-0.95, p=0.0032) on a larger microvascular composite outcome consisting of time to first occurrence of (i) albuminuria progression (as defined above), (ii) sustained decrease of ≥50 in eGFR (as opposed to 40 in the key endpoint), (iii) sustained ESKD, (iv) death due to kidney disease, (v) retinal laser coagulation or anti-VEGF injection for diabetic retinopathy, (vi) vitreous hemorrhage, or (vii) diabetes-related blindness. Notably, 745 of the 785 events in the linagliptin group and 810 of the 845 events in the placebo on this endpoint were albuminuria progression; with ~95% of events coming from one of seven outcomes, can this endpoint truly be considered a “composite microvascular” endpoint? Regardless, these are certainly intriguing results in the context of previous DPP-4 CVOTs. SAVOR-TIMI-53 for AZ’s Onglyza (saxagliptin) is the only other trial in the class to have prevented progression of microalbuminuria. To be sure, CARMELINA certainly provides evidence to support the idea that DPP-4s can prevent progression of albuminuria, though our sense is that the jury is still out on the impact on true microvascular complications (in that most don’t consider albuminuria a microvascular complication).

Hypoglycemia and Safety

Mark Cooper, PhD (Monash University, Melbourne, Australia)

Dr. Mark Cooper presented safety outcomes; consistent with the well-established safety profile of DPP-4s, no major adverse safety data was unearthed in the trial. Rates of adverse events (77% vs. 78%), adverse events leading to trial discontinuation (10% vs. 12%) and serious adverse events (37% vs. 39%) were well balanced between linagliptin treatment and placebo, respectively. Safety events of particular interest, including hypersensitivity reactions and renal adverse events, were also balanced between treatment and placebo. Rates of acute pancreatitis, although relatively low across the entire population (~0.2%), were higher in treatment than placebo (total events were 9 vs. 5; 0.3% vs 0.1% for linagliptin vs. placebo). Finally, rates of hypoglycemia were well balanced in the overall group (~20% for both), in insulin-treated participants (30% vs. 33%), sulfonylurea-treated participants (16% vs. 14%), and those with kidney disease (28% vs. 29%) for treatment vs. placebo, respectively. A similar well-balanced trend emerged for rates of severe hypoglycemia as well.

Summary, Conclusions, Clinical Perspectives, and Implications for Practice

Bernard Zinman, MD (University of Toronto, Canada)

According to Dr. Bernard Zinman, CARMELINA fills an important gap in data for the DPP-4 inhibitor class. Specifically, he pointed to the high proportion of patients with kidney disease (eGFR <60 ml/min/1.73 m²), comprising 63% of the total population. For comparison, 16% of patients in the SAVOR-TIMI trial for AZ’s Onglyza (saxagliptin) had an eGFR <50, 29% in the EXAMINE trial for Takeda’s Nesina (alogliptin) had an eGFR <60, and 9% in the TECOS trial for Merck’s Januvia (sitagliptin) had an eGFR <50. Additionally, CARMELINA had a markedly higher average proportion of participants with macroalbuminuria, clearly differentiating its population from that of other DPP-4 CVOTs (below left). Also separating CARMELINA from the pack is the fact that it found CV safety with the best trend for hospitalization for heart failure in the class to date (below right) – though we note that all trials apart from SAVOR-TIMI were statistically neutral on this endpoint. Continuing to renal outcomes, Dr. Zinman pointed to linagliptin’s significant reduction in albuminuria progression and the microvascular composite outcome as bright spots from the trial, and noted that CARMELINA is unique in the DPP-4 class inasmuch as its key secondary kidney outcome was pre-specified, adequately powered, and adjudicated. From our point of view this likely makes little clinical difference, as linagliptin was found to be renally safe compared to placebo – an opinion which most seem to already hold, from where we stand. Bringing things full circle, Dr. Zinman highlighted linagliptin’s reassuring safety profile despite the generally worse renal health of the study’s population – an important finding for a population that is often uniquely difficult to treat.

Commentator

Philip Home, DPhil (Newcastle University, UK)

In providing independent commentary, the great Prof. Philip Home first focused on what’s good about the study, namely:

• Linagliptin was established as safe, in the medium-term, in a high-CV-risk population.

• Consistency with other DPP-4 inhibitor CVOTs in lower-risk populations and for longer time periods is encouraging.

• There was no safety signal on heart failure.

• CARMELINA confirms the albuminuria reduction found with other glucose-lowering agents.

And then, Prof. Home discussed what could have been better about the study:

• Discontinuation from allocated therapy was 26% over 2.2 years.

• Investigation in two separate populations (i.e. some had only CKD, and some had only CVD) may not make sense or be easy to apply – these populations should be analyzed separately.

• There was a relatively low number of patient years of exposure (6,639).

• Most events were CV death, which doesn’t ring true with most diabetes patients.

• The microvascular endpoint is not truly a microvascular endpoint, as the effect was driven by albuminuria – which is a marker of vascular function, not microvascular complications. Moreover, he says, it’s not reasonable to expect renal outcomes to be impacted in such a short study of a drug that does not act on the kidney.

Oral Presentations: New Insights from Clinical Trials with Incretin-Based Therapies

Potential Impact of Differential Drop-In of Open-Label Diabetes Medications in EXSCEL

John Buse, MD (University of North Carolina, Chapel Hill, NC)

In a riveting oral presentation, the great Dr. John Buse presented a much-anticipated analysis of EXSCEL trial data, from the CVOT for AZ’s Bydureon: Did high rates of open-label drug-in diabetes medication use contribute to the trial’s very narrow miss on CV superiority (HR=0.91, 95% CI: 0.83-1.00, p<0.001 for non-inferiority, p=0.061 for superiority)? Over the past year, we’ve heard dozens of thought leaders assert that EXSCEL does far more to support cardioprotection as a class effect of GLP-1s than refute it – and also that EXSCEL in and of itself is a very positive trial. By and large, the trial’s statistical miss has been attributed to its pragmatic design, enrollment of a larger primary prevention population, use of complicated reconstitution kits, and lack of a run-in period to exclude those with low adherence (even just this last one could make results quite different from other trials that did this). Part of this pragmatic design involved patients continuing to see their usual care providers, who were allowed to prescribe any concomitant medications except another GLP-1 agonist (for context, LEADER protocol also prohibited DPP-4 inhibitors). It had been suggested that those in the placebo group of EXSCEL were prescribed, on average, more concomitant medications compared to the Bydureon group (we’re very curious how it would compare to other CVOTs, as well), tipping the scale in favor of placebo (below). Indeed, there were 36% (709) more antihyperglycemic medication drop-ins in the placebo group compared to the exenatide group (below). To account for this, researchers conducted a post-hoc analysis of the data, censoring patients at the time of drop-in for each of the six “open-label” medications considered (metformin, insulin, TZDs, SUs, GLP-1s, and SGLT-2s) and applying an estimated effect size for each mediation derived from published trials. However, Dr. Buse did note that this effect size is likely overestimated, because patients were censored at the time drop-in medication was added, then an estimated effect size derived from published trials for that agent was applied. Keeping this in mind, the hazard ratio on the three-point MACE shifted to nominal superiority when accounting for only sulfonylurea drop-ins (HR=0.89, 95% CI: 0.81-0.99, p=0.024), as well as for drop-ins of all six therapies (HR=0.87, 95% CI: 0.78-0.97, p=0.011) – certainly a positive result. The former analysis does raise some interesting questions, and though some may find it a bit strange that this would seem to indicate SUs confer some small degree of cardiovascular benefit, this is true short-term, but longer-term, SUs are associated with beta-cell burnout from what we understand. Big picture, of course, there are undoubtedly some confounding factors involved. That said, the confidence interval actually swung slightly in favor of placebo when only DPP-4s or GLP-1s were considered alone (below), leading Dr. Buse to note many limitations to the analysis. Principal among them, many adverse events were lost as patients were censored – events in the placebo and Bydureon groups dropped from 905 to 638 and 839 to 655, respectively, when all open-label medication drop-ins were considered. Nevertheless, we do see this analysis as another piece of evidence in support of the notion that many KOLs have shared with us - that EXSCEL very likely reflects cardioprotection with Bydureon.

• Notably, Dr. Buse pointed out that all concomitant GLP-1 prescriptions (182 in the exenatide group and 265 in the placebo group) violated the study protocol. This is certainly not an insignificant number, impacting 2.4% of the exenatide group and 3.6% of the placebo group. As we see it, this likely speaks to HCPs’ dedication to their patients’ wellbeing over a trial’s guidelines, reflecting the difficult trade-off placebo-controlled trials are beginning to present in the era of cardioprotective diabetes therapies. As such, the fact that this seems to have been such a big issue in EXSCEL perhaps suggests that future CVOTs should not be conducted with (i) a comparison arm that can’t receive the modern standard-of-care nor (ii) such “pragmatic” guidelines, so as to avoid pitting a what’s best for a patient against the integrity of a study. Dr. Buse suggested to us that future studies will need to have a pre-specified analysis plan to account for drop ins, which could also be an interesting viable alternative.

Near-Normoglycaemia, with Meaningful Discontinuations of Prandial Insulin, by Adding Weekly Albiglutide to Uncontrolled Basal/Bolus Insulin-Treated Type 2 Diabetes

Julio Rosenstock, MD (Dallas Diabetes and Endocrine Center, Dallas, TX)

Dr. Julio Rosenstock presented results from the HARMONY-10 SWITCH study (n=814), offering compelling evidence in support of the use of a GLP-1 agonist – specifically albiglutide – in addition to basal-bolus insulin regiments, and even as an alternative to prandial insulin in basal-bolus regimens. As a reminder, GSK announced plans to discontinue Tanzeum (albiglutide) in June 2017, but the once-weekly GLP-1 drew lots of attention at EASD 2018 with very positive results from the HARMONY-Outcomes CVOT (namely, a 22% reduction in three-point MACE). HARMONY-10 SWITCH is another component of the HARMONY program designed to evaluate Tanzeum’s potential as a prandial insulin replacement; the study took participants already on MDI and randomized to either (i) Tanzeum + Lantus (n=402) or (ii) Humalog + Lantus (n=412); both insulins were titrated to optimize FPG. At 26 weeks, A1c was non-inferior with the basal/GLP-1 regimen vs. the MDI regimen (6.7% vs. 6.6%, respectively), and both groups saw meaningful reductions from a shared baseline A1c of 8.2% – reflecting that optimization of MDI is critical, but that using a GLP-1 rather than prandial insulin can be as effective. Very notably, though, the basal/GLP-1 group saw a 10 mg/dl greater reduction in FPG vs. MDI, and the difference was significant (95% CI: -15, -5). Also positively, the difference between weight loss with Tanzeum vs. weight gain with Humalog was 4.4 kg (9.7 lbs), a very sizable and meaningful difference (95% CI: -4.9, -3.8), despite Tanzeum not being known to provide impressive weight loss. Moreover, the basal/GLP-1 regimen was associated with only 13 weekly injections, compared to 29 weekly injections with MDI. All of this said, Humalog was reintroduced in 18% of participants randomized to Tanzeum, though at a much lower dose (~10 units/day) than prior to randomization (the MDI group averaged ~70 units/day of Humalog at week 26). Our sense is that thought leaders are strongly in favor of using GLP-1s as an intensification to basal insulin (if not as the first injectable therapy, as the ADA/EASD consensus statement now recommends). As Dr. Rosenstock explained, GLP-1s hold a number of advantages over prandial insulin: They don’t cause weight gain or hypoglycemia, and a once-daily or once-weekly injection is much simpler than multiple, titratable insulin injections.

• The basal/GLP-1 group also saw a benefit on hypoglycemia: 57% of those on that regimen experienced severe or documented symptomatic hypoglycemia by week 26, compared to 75% of those on MDI (OR=0.43, 95% CI: 0.31-0.60, p<0.0001) – translating to a highly meaningful 57% reduced risk. As should be expected, MDI did have an edge on GI side effects; only 13% of participants experienced nausea, vomiting, or diarrhea with MDI, compared to 26% of those on basal/GLP-1. Accordingly, 3.5% of basal/GLP-1 participants withdrew from the study (primarily due to GI events), compared to 2.2% of those on MDI (primarily due to hypoglycemia).

Oral Presentations: GLP1 Receptor Agonists, SGLT2 Inhibitors and the Kidney: New Lessons from Large Clinical Trials

Semaglutide Treatment and Renal Function in the SUSTAIN 6 Trial

Robert Silver, MD (Southern New Hampshire Medical Center, Nashua, NH)

Dr. Robert Silver presented an intriguing post-hoc analysis of the very positive SUSTAIN-6 CVOT for semaglutide (Novo Nordisk’s Ozempic), demonstrating a significant decrease in albuminuria with semaglutide relative to placebo, regardless of baseline renal function. However, a contradictory and moderate decline in eGFR compared to placebo was observed in patients with normal renal function and mild renal impairment. For this analysis, participants from SUSTAIN-6 were grouped by baseline eGFR: normal (≥90 ml/min/1.73m²), mild (≥60 to <90 ml/min/1.73m²), moderate (≥30 to <60 ml/min/1.73m²), and severe (<30 ml/min/1.73m²). The primary composite endpoint of new or worsening nephropathy was examined, defined as new onset of persistent macroalbuminuria (>300 mg/g), persistent doubling of serum creatinine level and creatinine clearance ≥45 mL/min/1.73m², need for continuous renal replacement therapy, or death due to renal disease. Renal-related outcomes were change from baseline in eGFR and UACR. Semaglutide demonstrated dose-dependent positivity on the primary composite outcome of this post-hoc, as 1.0 mg of semaglutide gave a significant 48% reduction in new or worsening nephropathy vs. placebo (HR=0.52, 95% CI: 0.33-0.80, p<0.01) while the improvement conferred by the 0.5 mg dose was not significant (HR=0.76, 95% CI: 0.52-1.11) – see below. The trends for both doses of semaglutide were driven by significant reductions in persistent macroalbimunuria of 41% vs. placebo with 0.5 mg (HR=0.59, 95% CI: 0.37-0.93, p<0.05) and 50% with 1.0 mg (HR=0.50, 95%: 0.30-0.81, p<0.01).

• On eGFR, interestingly, semaglutide conferred variable results depending on renal function. Those with normal renal function or mild impairment experienced an initial, steep drop early on in the trial, followed by a stable, if slight, decrease in eGFR (essentially parallel with placebo) over the remainder. By the end of the 104-week trial, there was a significant drop in eGFR associated with 0.5 mg semaglutide vs. placebo (-10.4 vs. -8.4 ml/min/1.73m²) in those with normal renal function; the clinical relevance of this difference seems minor to us but this is speculation. Dr. Silver labeled the initial steep drop as “somewhat unexpected” but also emphasized the potential for semaglutide to confer long-term renal benefit, as average eGFR had stabilized and converged with placebo by the end – if the trial was extended, would placebo continue to fall below semaglutide? Indeed, the average eGFR of patients on semaglutide compared to placebo in the moderate and severe renal impairment groups were greater than those of placebo for the majority of the trial. Most notably, 1.0 mg of semaglutide conferred a significantly smaller decrease in eGFR compared to placebo at the end of the trial (-0.4 vs. -3.4 ml/min/1.73m²). This discrepancy in eGFR response to semaglutide is certainly puzzling, and Dr. Silver suggested that it could be indicative of a greater renal-protective effect as renal function continues to declines without treatment – an intriguing supposition.

• Both doses of semaglutide conferred significant decreases in albuminuria relative to placebo, as measured by UACR (urine albumin/urine creatinine), except for 0.5 mg semaglutide in those with severe renal impairment. This result did not appear to be dose dependent, nor did it appear to depend on the patient’s baseline renal function – contradictory to the data on eGFR. 

• Based on the results from this study, Dr. Silver hypothesized that semaglutide may have a beneficial renal-protective effect in patients with type 2 diabetes but conceded that a dedicated trial will be necessary to confirm as well as resolve the discrepancy in eGFR and albuminuria data in those with normal renal function and mild renal impairment. As we see it, this data is a bit too blurry as it stands to draw any concrete conclusions, though we would love to see a dedicated renal outcomes trial for every GLP-1. Evidence of renal-protection (as is emerging for SGLT-2s) would prove a major win for patients, outcomes beyond A1c, and the entire GLP-1 class.

Liraglutide Reduces Major Cardiovascular Events in Patients with Chronic Kidney Disease: Results from the LEADER Trial

Neil Poulter, MD (Imperial College London, UK)

Imperial College London’s Dr. Neil Poulter presented a new LEADER sub-analysis revealing that liraglutide’s cardioprotective benefits extend to people with type 2 diabetes and baseline renal disease. The findings were simultaneously published in Circulation. On the primary outcome of three-point MACE (CV death, nonfatal MI, and nonfatal stroke), LEADER participants with moderate to severe renal impairment (eGFR <60 mL/min/1.73 m², n=2,158, 23% of participants) actually saw a greater risk reduction with liraglutide than participants with milder renal impairment (eGFR ≥60 mL/min/1.73 m², n=7,182, 77% of participants): HR=0.69 (95% CI: 0.57-0.85) vs. HR=0.94 (95% CI: 0.83-1.07, p=0.01 for interaction). This pattern of numerically stronger risk reduction for people with more severe renal impairment persisted in the individual CV outcomes: On non-fatal MI, HR was 0.74 for participants with eGFR <60 (95% CI: 0.55-0.99) vs. an HR of 0.77 for those with eGFR ≥60 (95% CI: 0.77-1.13). On non-fatal stroke, HRs were 0.51 (95% CI: 0.33-0.80) vs. 1.07 (95% CI: 0.84-1.37). On CV death HRs were 0.67 (95% CI: 0.50-0.90) vs. 0.84 (95% CI: 0.67-1.05). Finally, on all-cause mortality, HRs were 0.74 (95% CI: 0.60-0.92) vs. 0.90 (95% CI: 0.75-1.07). For us, the consistent magnitude of these results reflects a compelling pattern; we imagine renal impairment serves as a strong marker of patients at very high risk for CV events, so it’s not surprising this group would see such strong benefit despite making up less than one-quarter of the LEADER sample. All that said, this suggestion of greater risk reduction for three-point MACE with lower eGFR was not supported by a subsequent analysis of the LEADER data using more granular eGFR subgroups (eGFR <30, 30 to <45, 45 to <60, 60 to <75, 75 to <90, and ≥90 mL/min/1.73 m²; p = 0.13 for interaction), nor another analysis examining eGFR as a continuous variable (p=0.61). Of course, splitting the sample into so many groups decreases the power available to observe such effects; for the same reason, the positive effects in the low eGFR cohort are all the more impressive, while the lack of significance in the group with renal function doesn’t mean those patients don’t benefit. Together, these findings indicate that liraglutide’s effect on MACE is not meaningfully influenced by baseline eGFR – rather, and importantly, liraglutide remains cardioprotective across risk levels. Dr. Poulter was careful to note that, as a post-hoc analysis, the results must be taken with a grain of salt, but overall we are very encouraged by the suggestion that liraglutide is at the very least no less CV protective for the particularly high-risk population of people with diabetes and Stage 3 or above CKD.

Lesser eGFR Decline with Dulaglutide Regardless of Weight Changes in People with Type 2 Diabetes and Moderate to Severe Chronic Kidney Disease (AWARD-7)

Katherine Tuttle, MD (Providence Health Care, Spokane, WA)

An analysis of the AWARD-7 trial comparing dulaglutide (Lilly’s Trulicity) to insulin glargine demonstrated that, in patients with type 2 diabetes and moderate to severe CKD, Trulicity was associated with less eGFR decline than insulin glargine, regardless of efficacy on weight loss. As a reminder, AWARD-7 randomized participants 1:1:1 to either Trulicity 0.75 mg, Trulicity 1.5 mg, or insulin glargine (Sanofi’s Lantus), all in combination with insulin lispro (Lilly’s Humalog). Results presented at ADA 2017 demonstrated comparable glycemic control between the two groups, but Trulicity had a significant benefit on both weight loss and hypoglycemia risk. This new analysis examined eGFR at 52 weeks and found that Trulicity was associated with stable eGFR, while Lantus was associated with a significant decline in eGFR over 52 weeks – see full renal results for AWARD-7 from ADA 2017. However, it must be considered that body weight changes may affect creatinine levels (through muscle mass changes), so GFR estimations through creatinine-based equations may be confounded. Fortunately, cystatin C is not affected by muscle mass changes, offering a more reliable measure of renal function in a person with changing body weight; calculating eGFR through cystatin C also offers strong evidence of lesser eGFR decline with Trulicity vs. Lantus. Moreover, no association was actually seen between body weight changes and changes in serum creatinine, serum cystatin C, or eGFR. These results effectively demonstrate that the lesser eGFR decline associated with Trulicity vs. Lantus is not related to body weight loss – a very encouraging result for patients and one that makes us all the more excited for full REWIND results at ADA 2019.

Cardiovascular Safety and Efficacy of Exenatide Once-weekly in Patients with Moderate Renal Dysfunction in the EXenatide Study of Cardiovascular Event Lowering (EXSCEL)

Adrian Hernandez, MD (Duke University, Durham, NC)

One year after the release of the EXSCEL results at EASD 2017, Duke’s Dr. Adrian Hernandez unveiled a sub-group analysis confirming that exenatide’s neutral impact on CV outcomes extends to patients with moderate renal dysfunction. As a reminder, EXSCEL enrolled participants with Stage 3 CKD, down to an eGFR of 30 ml/min/1.73 m². On the primary outcome of three-point MACE, exenatide had a neutral impact for the ~22% of the EXSCEL population (n=3,191) with eGFR <60 ml/min/1.73 m², giving an HR of 1.01 (95% CI: 0.86-1.19). This held true for both the subset of participants with Stage 3a CKD (eGFR 45-59 ml/min/1.73 m²) and the subset of participants with Stage 3b CKD (eGFR 30-44 ml/min/1.73 m²); in the former, the HR was 0.97 (95% CI: 0.79-1.20), while in the latter, the HR was 1.11 (95% CI: 0.84-1.47). There was no statistical heterogeneity between these groups, suggesting that exenatide’s impact on MACE was consistent across the spectrum of renal disease (p=0.51 for interaction). The same pattern appeared for individual cardiovascular outcomes, including  CV death, fatal and non-fatal MI, fatal and non-fatal stroke, all-cause mortality, hospitalization for heart failure, and hospitalization for acute coronary syndrome. Together, this body of evidence confirms that exenatide is no less safe, from a cardiovascular perspective, in people with renal impairment – not a surprising outcome given previous, similar post hocs, but nevertheless positive given the relative lack of effective glucose-lowering drugs available to people with diabetes and renal impairment.

Lixisenatide and Renal Outcomes in Patients with Type 2 Diabetes: A Post-Hoc Analysis of the ELIXA Trial

Marcel Muskiet, MD (Amsterdam University Medical Center, Netherlands)

A post-hoc analysis of the ELIXA trial suggested GLP-1 agonist lixisenatide offers renal-protective benefits. These results were also just published in Lancet Diabetes Endocrinology. In ELIXA participants with microalbuminuria randomized to lixisenatide, UACR fell an average of 6.4% from baseline over the course of the 108-week trial, vs. a 14.8% increase in UACR for those on placebo (p=0.05). Similarly, in participants with macroalbuminuria, UACR fell an impressive 31.6% with lixisenatide, compared to an 11% rise in UACR with placebo (p=0.0084) – potentially pointing to stronger renal protection in those with more advanced nephropathy. In addition to slowing the progression of UACR, lixisenatide was also associated with a 23% risk reduction for onset of macroalbuminuria in patients without macroalbuminuria at baseline (HR=0.77, 95% CI: 0.62-0.96, p=0.0174). However, on the “harder” renal endpoint of change in eGFR (generally considered the more meaningful indicator of renal function), lixisenatide-treated patients did not significantly differ from those on placebo, though it’s also important to note that ELIXA’s short (by CVOT standards) duration of 2.1 years (median) could have made it difficult to observe substantial changes in eGFR. Nevertheless, this signal does align with microvascular outcomes from LEADER, which identified a 16% reduction in time to first microvascular event (encompassing renal and ophthalmic adverse outcomes with liraglutide (HR=0.84, 95% CI: 0.73-0.97, p=0.02), driven entirely by a 22% reduction in renal outcomes (HR=0.78, 95% CI: 0.67-0.92, p=0.003). While our sense is that GLP-1 agonists have offered less compelling evidence for renal-protection than SGLT-2 inhibitors, we certainly believe this idea deserves further investigation – particularly given that glucose-lowering options for those with renal impairment are limited. If GLP-1s can offer both powerful glucose-lowering and renal protection to people with impaired kidney function, it would be a huge win for patients. It’s worth noting that standalone lixisenatide comprises only a fraction of GLP-1 agonist prescriptions, but these results also have positive implications for the clinical profile of lixisenatide/insulin glargine coformulation Soliqua – an incredibly efficacious (in terms of A1c-lowering) therapy that we continue to believe is terribly underutilized in diabetes treatment.

Oral Presentations: The Bottom Line: What's the Best Basal Insulin?

The Clinical Benefits of IDegLira in DUAL VII were Achieved While Using a Simple Regimen with Fewer Injections and Dose Adjustments Compared with Basal-Bolus Therapy

Eden Miller, MD (Diabetes Nation, Bend, OR, USA)

Presenting a new DUAL VII analysis, Dr. Eden Miller demonstrated 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. From a patient perspective (even a very heterogeneous group), this is not at all surprising. As a reminder, DUAL VII investigated Novo Nordisk’s Xultophy (insulin degludec/liraglutide fixed-ratio combination) vs. basal-bolus therapy with Lantus (insulin glargine) and NovoLog (insulin aspart). Results reported at ADA 2017 demonstrated Xultophy’s non-inferiority to basal-bolus on A1c lowering, but also indicated superiority on a number of secondary outcomes, including an impressive 89% reduction in severe or blood glucose-confirmed symptomatic hypoglycemia rate, not to mention ~2.1 lbs of weight loss vs. ~5.8 lbs of weight gain with basal-bolus. Dr. Eden Miller explained that, while the clinical benefits of Xultophy are well-known among the academic community, there are additional and equally meaningful benefits that Xultophy provides in terms of treatment complexity and convenience. In the DUAL VII study, the daily regimen for patients in the basal-bolus treatment group included the use of two pens, two to five daily injections, and two to five daily SMBG tests (one for each injection); we also know that, if a patient requires very high doses of basal insulin, number of daily injections can quickly climb even higher depending on which type of insulin/pen they use. Conversely, the Xultophy treatment arm only required one pen, one daily injection, and one SMBG test daily; we often write about the “added convenience” Xultophy offers patients over other injectable therapies, but putting it into these numbers reveals just how stark the contrast is. Considering the superior clinical profile Xultophy offers over basal-bolus regimens PLUS this tremendous improvement in patient experience, it’s hard to imagine many patients would choose the latter over the former – particularly considering the GI side effects caused by GLP-1s are smoothed out in the combination drug. Of course, any patient should be able to make an informed decision about their treatment regimen – any some may choose basal-bolus – but our feeling is that this often doesn’t happen, whether due to issues of cost, lack of information, or provider resistance to combination therapies. Dr. Eden also noted that the decrease in injection requirements with Xultophy translates to lower patient burden, as previous work has demonstrated a strong link between patients’ perception of treatment burden and the number of daily injection they must take. Patient burden was also especially high in the basal-bolus treatment group with respect to insulin dose adjustments during the study period: While both groups required a similar number of basal insulin adjustments on average (~17 total adjustments for both groups over 26 weeks), basal-bolus treatment required 200 bolus insulin adjustments, while Xultophy treatment – by its very nature – required zero such adjustments (see figure below). Burden was also far lower for the Xultophy treatment group in terms of total insulin dose: After 26 weeks, patients in the basal-bolus group took an average of ~84 units of total insulin per day, compared to only ~40 units for the Xultophy group (p<0.0001); basal dose alone was also higher in the basal-bolus group than the Xultophy group. All in all, this is amazing analysis though we are shocked by anyone who is surprised – insulin is one of the very hardest drugs in the world to dose, and it is incredibly variable day-to-day. Xultophy’s glycemic-dependent nature, as well as the stability of next-gen basal insulin and Victoza, are both hugely exciting.

Oral Presentations: Clinical Use of Insulin: What Works and What Doesn’t

Comparison of Canagliflozin and Liraglutide as a Replacement for Bolus Insulin in Type 2 Diabetes Patients Well-controlled by Basal-bolus Insulin

Naoki Kumashiro, MD (Toho University, Tokyo, Japan)

Dr. Naoki Kumashiro took the stage to present results from a small (n=40) comparative study demonstrating the efficacy of canagliflozin (J&J’s Invokana) and liraglutide (Novo Nordisk’s Victoza) as replacement for bolus insulin in patients who were well controlled with basal/bolus regimens. Forty patients with well-controlled type 2 diabetes on basal/bolus therapy (baseline A1c 6.7%) were randomized to replacing their bolus insulin with either canagliflozin (fixed at 100 mg/day) or liraglutide (increasing from 0.3 to 0.9 mg/day; doses in Japan are lower). After 24 weeks, A1c was adequately maintained compared to baseline: In the canagliflozin group, A1c dropped a nominal 0.1% from 6.8% to 6.7%; in the liraglutide group, A1c fell 0.2% from 6.4% to 6.2%. Rates of overall hypoglycemia were not different between groups, while rates of nocturnal hypoglycemia decreased in both groups when compared to baseline (exact numerical data was not presented here). No severe hypoglycemia events occurred in either group, which is certainly positive. We were pleased to see that researchers used CGM in this study to track glucose fluctuations, though the data indicated that these fluctuations did not meaningfully change in either group from baseline; we would love to see time-in-range comparisons between the groups and compared to baseline measurements. As expected with canagliflozin and liraglutide, body weight and BMI also tended to decrease in both groups when compared to baseline; however, exact numbers were not presented here. Dr. Kumashiro also noted that the total score for diabetes-treatment-related quality of life (DTR-QOL) significantly improved with both groups when compared to baseline; in our view, the lower treatment burden of an SGLT-2 inhibitor or GLP-1 agonist over bolus insulin is where much of the benefit of this switch is experienced by patients. However, while all domain scores improved in the canagliflozin group, no improvement was see in the liraglutide group in the domains of anxiety and dissatisfaction with treatment; we wonder if this can be attributed to the injection burden or even GI side effects. Nevertheless, these data add to a growing body of evidence supporting the effectiveness of SGLT-2 inhibitors and GLP-1s as a replacement for bolus insulin in controlling post-prandial glucose levels, and our sense is that clinical practice is reflecting these improvements more and more. 

Corporate Symposium: Glycaemic Control and Beyond: Shifting the Paradigm in Diabetes Care (Sponsored by Novo Nordisk)

GLP-1 Receptor Agonists: Assessing Patient Responses

Melanie Davies, MD (University of Leicester, UK)

In an illuminating talk on patient responses to GLP-1 agonists during a Novo Nordisk-sponsored session, Prof. Melanie Davies presented data to show that despite the GI side-effects common with GLP-1s, patients are generally very satisfied with the therapy. That said, she also reviewed within-class differences in both efficacy and patient satisfaction. As Prof. Davies aptly put it, “Treatment satisfaction is the whole package: It’s efficacy and tolerability, but it’s also the device and experience […] We can’t assume a patient is less satisfied with an injection than an oral, and we can’t assume that a daily medication is worse than a weekly one.” Prof. Davies presented data from a paper she co-authored with Dr. Jane Speight. Utilizing the Diabetes Treatment Satisfaction Questionnaire (DTSQ), which measures patient satisfaction with treatment efficacy, convenience, and flexibility (among other factors), Drs. Davies and Speight found that patient satisfaction with high-dose Victoza was actually significantly (p<0.05) greater than with Januvia; the difference was numerically but not significantly greater with low-dose Victoza. Importantly, both GLP-1 Victoza and DPP-4 Januvia are taken once-a-day (the former as an injection, the latter as a pill), so there’s no difference in frequency of dosing. Patients who switched from Januvia to either high-dose or low-dose Victoza experienced a significant increase in treatment satisfaction. In some ways, we were surprised to see these results: ~20%-25% of patients experience GI side-effects on a GLP-1, while DPP-4 inhibitors are known to be some of the most tolerable agents in our diabetes treatment toolkit. GLP-1s (thus far) are injectable, while DPP-4s are oral. On the other hand, GLP-1s are more efficacious. It’s important to see that this greater efficacy corresponds to a better experience for the patient.

• Notably, differences were also detected between GLP-1s. In particular, DTSQ scores were significantly (p<0.0001) higher with high-dose Victoza than with Bydureon (this was before the approval and launch of Bydureon BCise; see below for full results). On this point, we’d be curious for more details on what drove this sizable difference. For example, the Bydureon pen uses a 23 gauge, 5/16’ needle, compared to the 32 gauge, 4 mm needle used for the Victoza pen – might this drive greater patient satisfaction? Overall, we think these findings indicate that patient satisfaction can be complex and doesn’t necessarily follow the rules of convenience or tolerability that one might expect; we also think this is an important reminder to ask patients exactly what they want and are looking for in diabetes therapy. It’s not just about collecting one overarching “patient perspective,” but about appreciating each individual’s patient perspective. For many, a trade-off in tolerability, dose administration, or frequency of dosing is acceptable if it means improved efficacy. Prof. Davies did note that weight loss is not included in the DTSQ, so it may underestimate GLP-1 agonist satisfaction. There is also relatively little data on GLP-1 satisfaction vs. other therapy classes. Additionally, we would be intrigued to see more current results, given that this analysis was performed in 2012: Where would Trulicity and Ozempic fall? What will this mean for oral GLP-1 agonists?

• Prof. Davies postulated that adherence may start to improve as more patients start taking more efficacious therapies. Highlighting Ozempic in particular, Prof. Davies was positive about the jump forward in glucose- and weight-lowering efficacy, even compared to other GLP-1s. She detailed that, across the phase 3 SUSTAIN 1-5 trials, 58%-79% of participants on either Ozempic dose achieved an A1c <7% (significantly better at p<0.0001 than Januvia, Bydureon, and Lantus). Further, 37%-66% achieved weight loss ≥5%, and 25%-56% achieved the composite of A1c reduction ≥1% and weight loss ≥5%. Moreover, while “the higher you start, the farther you fall” on A1c and weight, the difference between Ozempic and its comparators was consistent across baseline A1c and BMI groups. Additionally, a 92-trial meta-analysis found that the efficacy of all GLP-1 agonists is fairly consistent across different baseline factors; there was some suggestion that South Asian patients benefit less (with a smaller sample size representing this population), as well as that younger patients may benefit more. However, BMI and duration of diabetes did not impact degree of benefit. This is all critical because, as Prof. Davies outlined, adherence remains a major challenge: In chronic disease overall, non-adherence averages at 50% after one year, 3/10 of patients will stop taking a prescription even before the first fill runs out, and this costs Europe ~$145 billion per year. Certainly, issues of affordability, convenience, and psychosocial dynamics also play a huge role in disrupting adherence to therapy, and more needs to be done to make taking medications less burdensome. We do think drug efficacy and being able to see meaningful benefit is important for patient adherence – and HCPs should strive to show patients data on how their meds are helping. All things considered, though, it’s one piece of a complex puzzle, and we’d love to see more done to improve access to the highly-potent therapies that are on the market.

Comparative Effectiveness of Long-Acting GLP-1 Receptor Agonists

Vanita Aroda, MD (Georgetown University, Hyattsville, MD)

In giving a rundown of the comparative effectiveness of long-acting GLP-1s, Dr. Aroda detailed the fundamental differences between different GLP-1 molecules (below). This discussion furthered our interest in HARMONY Outcomes, the CVOT for GSK’s discontinued GLP-1 agonist Tanzeum (once-weekly albiglutide), as well as REWIND, the CVOT for Lilly’s Trulicity (once-weekly dulaglutide). Results for HARMONY are set to be released and presented at 12 pm on Tuesday, while topline results for REWIND should be announced by the end of 2018. Will half-life/duration of action shake out to be a key factor in whether a GLP-1 agonist can demonstrate cardioprotection? Of course, the molecule with the longest duration of action – Bydureon (exenatide QW) – narrowly missed out on significance in the EXSCEL CVOT (HR=0.91; 95% CI: 0.83-1.00). However, this result has been widely attributed to the trial’s large secondary prevention cohort (27%) and pragmatic design, which stands in contrast to 19% and 17% primary prevention cohorts in LEADER for Victoza and SUSTAIN 6 for Ozempic – see a side-by-side comparison of GLP-1 agonist CVOTs here. HARMONY and REWIND will shed important light on how baseline CVD status of enrolled patients impacts ability to demonstrate benefit: While the former enrolled 100% participants with baseline CVD (secondary prevention), REWIND has a very large 59% primary prevention cohort – and a longer duration of follow-up to compensate to the lower event rate. To date, the role of GLP-1 agonists for primary prevention remains sharply debated, primarily due to the lack of evidence, and we’re very much looking forward to REWIND results.

• As Dr. Aroda explained, these different pharmacokinetic profiles of GLP-1s have a big impact on their clinical utility. For example, Byetta and Adlyxin are very short acting and have strong effects on postprandial glucose, while the longer-acting but still once-daily Victoza has a stronger overall glucose-lowering effect and carries greater weight loss potential. Moving to once-weekly therapies, Bydureon offers similar efficacy with greater tolerability, Trulicity an improvement in glucose-lowering efficacy with once-weekly dosing, and finally Ozempic with even greater glucose-lowering and weight loss efficacy, as well as CV risk reduction. In Dr. Aroda’s assessment, not all therapies, nor all GLP-1 agonists, are created equally; in many ways, this allows for more flexibility in matching therapies to patient profiles and pathophysiology. But the biggest question we have now is: Is cardioprotection a class effect of GLP-1 agonists, or does cardioprotection vary based on some combination of these inherent molecular differences? With two more CVOTs in the class reading out before year-end, we expect this to be a big topic of discussion for the next few months.

  • Whether a GLP-1 agonist is based on human-GLP-1 vs. exendin has also been widely discussed as a potential mediator of CV effects. Dr. Jorge Plutzky touched on this issue in his subsequent discussion of cardioprotective mechanisms for the class, detailing that the molecules for Victoza, Ozempic, Trulicity, and Tanzeum are all human-based, while Bydureon, Byetta, and Adlyxin are all exendin-based. That said, even the human-based molecules feature significant differences: For example, semaglutide (Ozempic) is 4.11 kDa, while dulaglutide (Trulicity) is 63 kDa. Does a molecule’s size impact its ability to reach certain GLP-1 receptors throughout the body?

SGLT Inhibitors

Symposium: EASE Programme: Empagliflozin as an Adjunct to Insulin in Type 1 Diabetes

Efficacy of Empagliflozin

Julio Rosenstock, MD (Dallas Diabetes Research Center, Dallas, TX)

Results from the phase 3 EASE program for Lilly/BI’s Jardiance in type 1 diabetes – comprised of EASE-2 (n=730) and EASE-3 (n=977) – identified significant A1c, time-in-range, and weight loss benefits with three doses of empagliflozin (2.5 mg, 10 mg, 25 mg) – see the full paper, just published in Diabetes Care, and watch the webcast here. However, results also indicate a consistent dose-response relationship, in line with two other phase 3 programs of SGLTs in type 1 (AZ’s DEPICT program and Lexicon/Sanofi’s inTandem program). Following announcement of very topline results at ADA 2018 (i.e. no numbers), all eyes were on the 2.5 mg empagliflozin dose investigated in 241 EASE-3 participants only: According to Dr. Julio Rosenstock’s ADA presentation, this dose gave significant improvements on A1c and body weight with no increased risk of DKA – we’ve taken a deep dive into EASE safety data in the highlight below this. Dr. Rosenstock returned to the stage at EASD to present a truly enormous amount of data from both EASE trials, which we’ve summarized in the table below. At a high level, these results are consistent with other phase 3 trials of SGLTs in type 1. Across all three SGLT-in-type 1 programs, higher doses seem to offer considerably higher efficacy, but at the apparent cost of greater DKA risk – more on this one highlight below.

EASE-2 and EASE-3 Efficacy Results

	EASE-2			EASE-3
Endpoint (26 weeks)	Placebo (n=243)	Empa 10 mg (n=243)	Empa 25 mg (n=244)	Placebo (n=241)	Empa 2.5 mg (2=241)	Empa 10 mg (n=248)	Empa 25 mg (n=245)
A1c Reduction (placebo-adjusted)	-	0.54% (95% CI: -0.65, -0.42; p<0.0001)	0.53% (95% CI: -0.65, -0.42; p<0.0001)	-	0.28% (95% CI: -0.42, -0.15; p<0.0001)	0.45% (95% CI: -0.58, -0.32; p<0.0001)	0.52% (95% CI: -0.66, -0.39; p<0.0001)
Weight Loss	-	2.7 kg (6.0 lbs) (95% CI: -3.3, -2.1; p<0.0001)	3.3 kg (7.3 lbs) (95% CI: -3.8, -2.7; p<0.0001)	-	1.8 kg (4.0 lbs) (95% CI: -2.3, -1.2; p<0.0001))	3.0 kg (6.6 lbs) (95% CI: -3.6, -2.5; p<0.0001)	3.4 kg (7.5 lbs) (95% CI: -4.0, -2.9; p<0.0001)
Time-in-Range Increase	-1% (-14 minute)	+10% (+2.4 hours)	+12% (+2.9 hours)	+2% (+29 minutes)	+6% (+1.4 hours)	+12% (+2.9 hours)	+8% ( +1.9 hours)
Percent Change in Total Daily Insulin Dose		13% (p<0.0001)	13% (p<0.0001)		6% (p<0.0001)	10% (p<0.0001)	13% (p<0.0001)
Systolic Blood Pressure (placebo adjusted)		-2.1 mmHg (NS)	-3.7 mmHg (p<0.001)		-2.1 mmHg (p<0.001)	-3.9 mmHg (p<0.001)	-3.7 mmHg (p<0.001)

• An analysis of A1c reduction by baseline A1c found numerically greater benefit in those with a starting A1c ≥8.0%. Among the ~60% of participants in this group, the 2.5 mg dose gave a higher 0.35% A1c drop, vs. 0.53% with 10 mg and 0.62% with 25 mg. All of these were significant at p<0.0001 vs. placebo. For those with baseline A1c <8.0%, a 0.18% drop with 2.5 mg empagliflozin was not significantly, but a 0.35% drop with 10 mg and a 0.39% drop with 25 mg were both significant vs. placebo (p<0.001). This furthers a point of discussion we’ve heard elsewhere: Should SGLTs be initiated in patients with type 1 and higher A1cs? We don’t imagine many would balk at using an SGLT-2 in a patient with an A1c of 8.2%, for example, but some have expressed concern that using these in those with very sub-par glycemic control would expose them to a very high risk of DKA. On the other hand, this analysis supports the notion that those patients could actually benefit the most. Our sense is that the field has not reached a clear consensus of these aspects of patient selection.   

  • In presenting efficacy results, Dr. Rosenstock strongly emphasized that “2.5 mg is efficacious in those with A1c over 8% at baseline. In our view, however, the lack of efficacy in those who already have good glucose control may be a sticking point for this dose (though a larger sample size may have demonstrated a significant effect). Clearly, there are some concerns about the fairly substantial drop in efficacy from 10/25 mg to 2.5 mg and aren’t convinced this will hold clinically meaningful benefit for most patients, but we’re also very reassured by the safety data. Could a 2.5 mg dose of empagliflozin be easier for HCPs to prescribe and patients to take? Or should there just be recommendations that reduce the DKA risk for all SGLTs? Clearly, substantial safety protocols should be in place regardless – and while EASE-3 has raised interesting questions on this front, we do come back to the fact that patients want better glucose control and so too reduced efficacy will probably not make this worthwhile for a decent range of patients.

  • We would love to see all SGLT inhibitors investigated for cardiovascular benefit in people with type 1 diabetes. There’s an enormous unmet need for therapies to lower the higher cardiovascular risk that people with type 1 diabetes face, and we think it’s very possible that the most meaningful benefits SGLTs will offer patients with type 1 are not actually through glucose lowering, but rather cardio- and renal-protection. 

• Study design and baseline characteristics: Both EASE-2 and EASE-3 were designed with six weeks of insulin optimization and a two-week placebo run-in period (with CGM) prior to randomization. In EASE-2, participants were randomized equally to placebo, 10 mg empagliflozin, or 25 mg empagliflozin, and CGM was used for weeks 23-26 and week 52. The primary endpoint was at 26 weeks, followed by another 26 weeks of treatment plus three weeks of post-study follow-up. In contrast, EASE-3 randomized patients equally to placebo or one of three empagliflozin doses (2.5, 10, or 25 mg). EASE-3 was only 26 weeks and included CGM at the start and end of the trial, as well as three weeks of post-study follow-up. All participants were adults with type 1 for ≥one year, on stable MDI or CSII, with an A1c ≥7.5% or ≤10% and eGFR ≥30 ml/min/1.73 m². Severe hypoglycemia or DKA in the prior three months excluded any participant. EASE-2 and EASE-3 were well-balanced on baseline characteristics, including age (45 and 43 years), BMI (29 and 28 kg/m²), time since diagnosis (23 and 21 years), mean A1c (8.1% and 8.2% after insulin optimization), and insulin delivery (59% and 66% MDI; 41% and 34% CSII), all respectively.

  • We wonder why EASE-3 was not also extended to 52 weeks; we might speculate that Lilly/BI were in a rush to finish the phase 3 program, given two others in this area have completed. Mostly importantly, we think it would be very valuable to have longer-term data on the 2.5 mg dose. We’re not sure how heavily this will be considered a limitation of EASE’s data.

  • As background, the design to investigate a 2.5 mg dose of empagliflozin was based on the phase 2 EASE-1 study led by Dr. Thomas Pieber, which found that urinary glucose excretion was largely maintained (at ~73 g/day) with 2.5 mg in people with type 1 diabetes. The 10 mg dose gave ~103 g/day of UGE and the 25 mg dose ~102 g/day – so, virtually identical. While people with type 2 diabetes see a steady decline in UGE as the empagliflozin dose decreased, it seems like there might be an upper limit on UGE among those with type 1; empagliflozin 2.5 mg comes close to this limit. In attempting to explain this phenomenon, Dr. Rosenstock cited greater glucose variability (i.e. more opportunities to spill glucose), higher renal filtration rate, and greater expression/activity of SGLT-2 as potentially combining for the higher urinary glucose excretion in type 1 diabetes. Based on these results – including comparable A1c reductions across all three doses in phase 2 – the phase 3 program was designed.

Close Concerns’ Questions

• Will AZ and Sanofi/Lexicon consider investigating lower doses of Farxiga and Zynquista, respectively?

• What are the clinical implications of the 2.5 mg results, in terms of how easy it will be for prescribers and patients to adopt empagliflozin?

• How will regulatory agencies consider the risk/benefit ratio of different SGLT inhibitor doses?

Safety Profile and Hypoglycaemia

Bruce Perkins, MD (Leadership Sinai Centre for Diabetes, CA)

Dr. Bruce Perkins presented much-anticipated safety data for the EASE-2 and EASE-3 studies, highlighting the absence of increased DKA risk with the 2.5 mg dose of empagliflozin. In EASE-3 over 26 weeks, there were three certain DKA events among 241 participants randomized to 2.5 mg (1.2%), compared to two certain events in the placebo group (0.8%). Dr. Perkins provided a thorough analysis of data pertaining to general safety, adverse events, and DKA. Methodology wise, data sets for 10 mg and 25 mg empagliflozin treatment groups were pooled separately across the two studies to generate a larger data pool; placebo groups were also pooled. Separately, the 2.5 mg empagliflozin arm from EASE-3 was compared to the placebo group from EASE-3 only. Broadly speaking, there was a dose-dependent response in DKA risk inasmuch as DKA risk was increased with both the 10 mg and 25 mg doses, but not with the 2.5 mg dose; however, there were actually more confirmed DKA events in the 10 mg group (21 events; 4.3%) than the 25 mg group (18 events; 3.3%), though both were meaningfully higher than placebo (6 events; 1.2%). The higher doses were also generally associated with elevated rates of other adverse events, including genital infections, volume depletion, and DKA. This dose-response relationship for risk is generally in line with what has been observed in other trials of SGLTs in type 1.

• Looking past “certain DKA” to “potential DKA” the proportion of patients with potential DKA or potential ketosis both trended towards a higher fraction in the 2.5 mg group when compared to placebo (3 events/1.2% vs. 1 event/0.4%, and 7 events/2.9% vs. 2 events/0.8%, respectively), indicating a potential residual risk for DKA even with lower doses of treatment.

• There was one death in the 25 mg empagliflozin arm due to DKA, which to our understanding is attributable to empagliflozin. There is a write up of the clinical case in the paper’s supplementary materials, which serves as a poignant reminder of just how serious this complication can be. Moreover, we’re more convinced than ever that, if approved, SGLT use in type 1 needs to be accompanied not only by extensive patient and provider education but also a widescale effort to make HCPs and other stakeholders more aware of this risk, what it looks like, and how to treat it.

• A few sizable limitations of this data (the 2.5 mg finding in particular), in our view, are that (i) the sample size is small, (ii) the event rate overall is low, and (iii) this is a single trial. We do think there’s an important historical lesson to keep in mind: When DEPICT 1, one of two phase 3 trials for AZ’s Farxiga in type 1, read out at EASD 2017, there was no real signal for DKA. However, almost one year later when DEPICT 2 was presented at ADA 2018, we learned that an imbalance had been identified in this second, nearly-identical study. In contrast, all three phase 3 trials for Sanofi/Lexicon’s sotagliflozin in type 1 (inTandem 1, inTandem 2, and inTandem 3) have identified an increased and apparently dose-dependent DKA risk with the SGLT-1/2 dual inhibitor. Indeed, our sense is that many already considered DKA a class effect of SGLTs in type 1 diabetes despite the reassuring data from DEPICT 1. All of this being said, a 2.5 mg dose of empagliflozin is almost certainly the most “mild” SGLT inhibition to be investigated to date in a phase 3 program; this represents one-quarter of the lowest dose approved for type 2 diabetes. The DEPICT program investigated the two approved doses of Farxiga (5 mg and 10 mg), while Lexicon/Sanofi investigated 100 mg and 200 mg sotagliflozin. As such, there’s very little to compare this finding to, and we expect significant discussion as to whether a single 241 person arm with only five “certain” events (three in the EASE-3 placebo arm and two in the 2.5 mg arm) is enough evidence to draw strong conclusions on low-dose safety. Moreover, the efficacy trade-off seriously complicates the discussion.

• Intensive ketone monitoring procedures were set during the placebo run-in and the first four weeks of treatment, as patients were asked to perform daily beta-hydroxybutyrate (BHB) measurements to establish baseline ketone levels. For the remainder of the trial, this recommendation was eased to only two to three measurements per week, or in situations where clinical symptoms of DKA were present. The face that this level of monitoring/checking was implemented and an increased incidence of DKA still occurred is meaningful to the ongoing debate about ketone monitoring when using SGLTs for type 1: How can this procedure be more effective? Complicating things further, these monitoring requirements represent a significant burden on the patient (in terms of both cost and effort), and we are curious about how patients would receive these recommendations in real-world settings (thought leaders also remain conflicted on this topic of ketone monitoring frequency).

  • DKA events were adjudicated upon certain trigger events: An investigator reporting symptoms related to DKA, even if absent of a corresponding BHB measurement; all BHB measurements between 1.5-3.8 mmol/l accompanied by clinical symptoms of DKA; and any BHB measurement greater than 3.8 mmol/l. These cut points arbitrated by study investigators generally align with those agreed upon by thought leaders, as gleaned by conversations our team has had, but nevertheless still remain a point of discussion. A total of 196 adjudicated cases emerged over the two trials, with 86 of these determined to be DKA events. Of these 86 cases, nine were classified as severe (defined as pH <7 or bicarbonate <10 mEq/L or stupor/coma), 25 as moderate (pH between 7-7.24 or bicarbonate between 10-15 mEq/L), and the remaining 52 cases as mild (pH between 7.25-7.30 or bicarbonate between 15-18 mEq/L or no neurological symptoms).

  • Euglycemic DKA (euDKA) events were only present in treatment groups and did not occur in placebo. Nine out of the 21 DKA events in the 10 mg group were classified as euDKA (<250 mg/dl), while five out of 16 in the 25 mg group and one out of two in the 2.5 mg group were also classified as euDKA. No such events occurred in placebo. The paradoxical nature of euDKA presents a significant challenge in clinical practice and patient education, as it complicates the recognition and diagnosis of DKA and can delay proper and effective treatment. These safety data clearly establish euDKA as a risk with SGLT-2 inhibitors; the question now becomes how educational efforts with HCPs and patients can better mitigate this risk.

• DKA risk was increased in women in the empagliflozin treatment group, while no difference was seen in DKA rates for men between placebo and treatment groups. Women in the 10 and 25 mg groups had rates of certain or potential DKA of 10.5% and 8.7%, respectively, compared to 3.1% in the placebo group. On the other hand, men showed no such effect: Rates of DKA were 2.6% and 2.5% for 10 and 25 mg groups, respectively, compared to 3.1% for placebo. This is certainly an interesting finding, and we are currently trying to learn more about why this may be, as well as if other trials have observed a similar effect.

• Insulin pump use was also associated with higher risk of DKA. Patients using CSII in the 10 mg and 25 mg treatment groups had rates of potential or certain DKA of 12.6% and 9.9% respectively, compared to 3.9% for placebo. Conversely, those on MDI had more balanced (and lower) rates of DKA, albeit still higher in treatment groups: 3.2% for both 10 and 25 mg groups compared to 1.3% for placebo. These results further reinforce pump failure as a major risk factor for DKA.

• Hypoglycemia: Hypoglycemia data was collected both as investigator-reported events (n=~13,000) and as patient-reported events (n>23,000). Describing the distinction between these two categories, Dr. Perkins noted that while investigator-reported events are inherently influenced by subjective judgement, patient-reported events consider the totality of data and also encompass the investigator-reported events as a result. Generally, investigator-reported events did not find a difference between hypoglycemic events in treatment and placebo; however, patient-reported events did show a reduction in hypoglycemic events, including a reduction in nocturnal hypoglycemia events, across all doses of empagliflozin treatment. The screenshots below show point estimates and confidence intervals for this data; of particular note is the consistent trend seen in the patient-reported event data favoring empagliflozin treatment. We’re looking forward to learning more about the relative validity and meaning or investigator-reported vs. patient-reported events.

 

 

Translating Evidence to Practice

Lori Laffel, MD (Joslin Diabetes Center, Boston, MA)

Following presentation of safety and efficacy data for the phase 3 EASE program of Jardiance in type 1 diabetes, Dr. Lori Laffel discussed clinical implications of these results. What are the next steps for applying EASE results to clinical practice? Dr. Lori Laffel dove into this translational topic, stressing that mitigation of safety risks – especially DKA – will be the bottleneck on how successful and widely applicable this therapy can become. All stakeholders must be properly educated in order for these agents to be used safely in type 1 patients, including the patients themselves but also family members, HCPs, payers, emergency room staff, and others – the question, then, is how to implement such largescale education? Notably, Dr. Laffel underscored the role that payers will have in this environment, as they will need to provide BG/CGM supplies along with blood ketone strips and meters for more frequent testing of this important marker of ketosis. In our view, this provision is by no means guaranteed (though we think it should be), and remains a significant barrier to safe use – ketone strips are costly, and many patients would likely forgo blood ketone testing due to cost.

• The STICH protocol should also be communicated to patients, and Dr. Laffel touted the use of wallet cards as an effective education and safety measure (shown below). Patient selection will also be key: Similar to thoughts we’ve heard before from other KOLs, Dr. Laffel recommended steering away from SGLT-2 inhibitor use in certain patient populations, including those with eating disorders, those on low-carb diets, and those with a history of recurrent DKA. Finally, Dr. Laffel also recommended “the lowest dose that balances efficacy with safety” in terms of DKA risk minimization. However, it’s hard to answer this question: Does the efficacy offered by 2.5 mg empagliflozin meet the threshold we’d want from an adjunct type 1 treatment? Efficacy will likely mean different things to different people, and Dr. Laffel left the question open in closing: Is the 2.5 mg dose a “sweet spot” for balancing efficacy and risk?

Commentator

Thomas Pieber, MD (University of Graz, Austria)

Following presentation of safety and efficacy data for the phase 3 EASE program of Jardiance in type 1 diabetes, Dr. Thomas Pieber presented an engaging commentary comparing the risk/benefit profile of empagliflozin to other SGLT inhibitors studied in type 1. Dr. Pieber also commented on the future action needed to fully realize the potential of this class in type 1 patients, drawing heavily upon a recent meta-analysis of 14 studies investigating SGLT-2 inhibitor use in type 1 patients, adding comparison to EASE results.

• Benefits of empagliflozin: Dr. Pieber contextualized the “clear benefits” of A1c reduction, weight loss, and insulin dose reduction observed with empagliflozin in EASE. In terms of A1c, empagliflozin treatment effects build nicely upon previous results with the SGLT-2 inhibitor class in type 1 patients. The abovementioned meta-analysis found a weighted mean A1c reduction from placebo of -0.40 (95% CI: -0.46 to -0.35); comparatively, empagliflozin 10 and 25 mg treatment gave an average A1c reduction of ~0.50 in EASE-2 and EASE-3. Notably however, the lower 2.5 mg dose of empagliflozin dovetailed from this trend with its 0.28% reduction in A1c, reflecting the generally decreased efficacy offered with this dose. The benefits in weight loss and insulin dose reduction seen in the EASE program also aligned well with those seen in previous studies, save the 2.5 mg dose.

• Risks of empagliflozin: Meta-analysis shows an overall increased risk of genital tract infection with SGLT-2 inhibitor in those with type 1 (HR=3.44, 95% CI: 2.34-5.07). Again, EASE data is aligned with this range, with hazard ratios of 3.48 and 3.55 for the 10 and 25 mg doses. As noted before, these safety concerns were attenuated with the lower dose, which showed a reduced risk at HR=1.65 (significance was not clear). A similar trend was seen with DKA: Overall risk from the meta-analysis (HR=3.38, 95% CI: 1.76-6.56) aligned with DKA risk at the higher doses of empagliflozin (HR=3.58 for 10 mg and HR=2.75 at 15 mg), but not the 2.5 mg dose, which was again lower (HR=0.66).

• Future action: Dr. Pieber advocated for several step to be taken prior to approval of any SGLT inhibitor for adjunct therapy in type 1 diabetes. He urged investigators to study more doses of empagliflozin – perhaps at 5 or 7.5 mg – in order to see if increased glycemic and weight benefits could be achieved at these doses while still avoiding increased risk of DKA. We’re particularly intrigued by this idea and would love to see Lilly/BI, as well as other manufacturers, study more doses, which they said they would be doing on Lilly’s investor call this week (15 mg is the highest dose). Dr. Pieber also called for mechanistic studies to better understand the progression of DKA risk in relation to the SGLT-2 inhibitor mechanism. According to him, these studies have already been requested by FDA and EMA. Interestingly, Dr. Pieber also called for CVOTs of SGLTs in type 1 patients, a sentiment we’ve heard expressed more frequently in recent times, including by Dr. David Cherney in the final talk of ADA 2018.

Symposium: Comparative Effectiveness of Cardiovascular Outcomes in New Users of SGLT-2 Inhibitors (CVD-REAL)

Comparative Effectiveness of Cardiovascular Outcomes in New Users of SGLT-2 Inhibitors (CVD-REAL)

Kamlesh Khunti, MD, PhD (Leicester Diabetes Center, UK), Matthew Cavender, MD (University of North Carolina, Chapel Hill, NC), Kare Birkeland, MD, PhD (University of Oslo, Norway), Mikhail Kosiborod, MD (Saint Luke’s Mid-America Heart Institute, Kansas City, MS), Ele Ferrannini, MD (University of Pisa, Italy)

An exciting session on CVD-REAL (AZ’s real-world investigation of CV benefit with SGLT-2 inhibitors) brought with it expanded data from CVD-REAL Nordic and a fascinating Q&A on the clinical implications of real-world-data (RWD). Notably, the esteemed presenters of this session all advocated for wider use and acceptance of RWE in clinical, regulatory, and industrial decisions. We certainly agree that RWD studies – especially when as robust as CVD-REAL – are very valuable additions to CVOT and other RCT data, allowing the field to generate a more complete picture of a therapy’s effects for patients and providers. Dr. Kamlesh Khunti echoed this sentiment in his attempt to integrate RWD and RCT evidence in favor of SGLT-2 cardioprotection. As he sees it, the greater generalizability of the patient population in real-world data coupled with the fact that trends on three-point MACE and hospitalization for heart failure all match those of RCTs allow RWD (such as CVD-REAL) to extend findings from CVOTs to populations not represented in these trials. For his part, Dr. Mikhail Kosiborod underscored the holistic impact of CVD-REAL, noting that these trials come at a fraction of the cost of RCTs yet also have a significant impact on clinical practice and a growing impact on regulatory decisions. On the former, Dr. Kosiborod shared the results of a post-conference survey question from ACC 2018 (where CVD-REAL 2 was presented): CVD-REAL 2 had the greatest impact on clinical practice and future research agendas among attendees when put up against a CANTOS diabetes analysis (anti-inflammatory canakinumab’s effect on incident type 2 diabetes) and a clinical focus session on the new reality of complete diabetes care. On the increasing importance of RWE for regulators, Dr. Kosiborod highlighted a recent JAMA publication by FDA outlining its viewpoint for using RWD in regulatory decisions. According to the article, under the 21^st Century Cures Act, FDA is tasked with developing a program to evaluate the use of RWE to support approval of new indications for approved drugs or to satisfy post-approval study requirements (CVOTs, perhaps?). The article notes that a framework for this program will be published by the end of 2018, which will hopefully be informed by the recent ADA Symposium on the use of RWD to improve the prevention and care of diabetes-related outcomes. Finishing out the session was independent commentator Dr. Ele Ferrannini, who suggested that it would be silly not to use the information from both RCTs (for hypothesis testing) and RWE (for healthcare impact), seeing as they have complementary strengths. Specifically, he pointed to the randomization, thorough monitoring, and endpoint adjudication of RCTs as lacking in generalizability, data on rare adverse events, and long follow-ups – all strong aspects of RWE.

• University of Oslo’s Dr. Kare Birkeland shared expanded data from CVD-REAL Nordic – specifically, a sub-analysis of participants starting new treatment with AZ’s SGLT-2 inhibitor Farxiga (n=10,227) or any DPP-4 inhibitor (n=30,681). Dr. Birkeland reported a significant 21% reduction on three-point MACE (non-fatal MI, non-fatal stroke, CV death) favoring dapagliflozin over DPP-4s (HR=0.79, 95% CI: 0.67-0.94, p=0.006), as well as an impressive 38% reduction on hospitalization for heart failure (HR=0.62, 95% CI: 0.50-0.77, p<0.001). Turning to the individual components of MACE, hazard ratios were non-significant but still favored dapagliflozin: 0.91 for non-fatal MI (95% CI: 0.72-1.16, p=0.445), 0.79 for non-fatal stroke (95% CI: 0.61-1.03, p=0.086), and 0.76 for CV death (95% CI: 0.53-1.08, p=0.122). Hazard ratios for MACE plus unstable angina (HR=0.81, 95% CI: 0.69-0.94, p=0.007), MACE plus unstable angina and hospitalization for heart failure (HR=0.75, 95% CI: 0.66-0.86, p<0.001), and all-cause mortality (HR=0.59, 95% CI: 0.49-0.72, p<0.001) were all found to be significant. Severe hypoglycemia events did not differ significantly between the two groups, which Dr. Birkeland characterized as “not surprising.” While CVD REAL Nordic results using registries in Sweden and Norway were first presented at EASD 2017, this analysis (published in Diabetes, Obesity and Metabolism) included data from the Denmark patient registry, adding 15,460 patients. Prior CV disease was identified in ~23% of patients overall. While there is always inherent risk of confounding with real-world studies, we think the comparison of SGLT-2s to DPP-4s is likely more robust than comparing SGLT-2s to any other class of medication – that is, it seems likely that patients receiving SGLT-2s vs. DPP-4s likely have more baseline factors in common than those receiving SGLT-2s vs. sulfonylureas, for example. Moreover, the confirmation of CVOT evidence, in this case, is all the more compelling. However, it is notable that Farxiga gave a significant effect on three-point MACE in this registry-based dataset but not in the DECLARE CVOT.

Panel Discussion

Q: Most guidelines include SGLT-2 inhibitors ahead of DPP-4 inhibitors, but DPP-4 inhibitors are safer. How do you reconcile that? What about metformin?

Prof. Khunti: There’s a lot of debate on metformin – should SGLT-2 be the first line because of the data coming in now. For metformin, it’s mostly the data from the UKPDS that showed benefit. In meta-analysis, there doesn’t seem to be an association of metformin as first-line therapy, as much as in UKPDS. There is consensus, some authorities are already saying this, that we should at least consider SGLT-2 as first-line in patients with established cardiovascular disease, some momentum gaining there.

Prof. Kosiborod: What your commenting on is the CV and renal effects for SGLT-2 inhibitors, but the more consistent safety with DPP-4 inhibitors. There is some very important data on safety for SGLT-2 inhibitors coming out of the DECLARE study. If you recall, it is the largest study of SGLT-2, as many enrolled as EMPA-REG OUTCOME and CANVAS combined, and also the longest follow-up at 4.5 years – a year longer than median follow-up in previous trials. Both of those are important from safety standpoints. I think that additional data will be incredibly critical to understand safety, not just efficacy, and overall value proposition when you compare benefit-risk. It’s important, remember that in clinical practice, we don’t just use medications because their safe, we use them because of the balance between benefit and safety. In cardiology, we prescribe medicines all the time that have real risks, but also good benefits – if the benefits outweigh the risks, then we prescribe it. If all we wanted to do was give 100% safety, then we’d just give placebo. We don’t do that in medicine. I think we’ll be a lot better informed on SGLT-2 safety and efficacy balance after DECLARE.

Prof. Khunti: I’m not sure I’d give placebo either if we wanted 100% safety, because 20%-30% of the placebo arm in some studies still have adverse events!

Q: In your presentation, it is apparent that real-world evidence (RWE) complements data from the clinical trial. Where do we go if the RWE contradicts the RCT? How do we assess that paradox?

Prof. Khunti: That’s what Prof. Ele Ferrannini asked in his commentary in the main symposium. You need to look at the balance; that’s why we do meta-analyses for RCTs. You need to see the direction of travel in the effect size: if you see a consistent direction of travel – we’ve seen that in whatever SGLT-2 you use – then you can be confident that you’re qualitatively in the right direction. If you’re going the wrong way, then I think you need to think about waiting for further RWE. For example, a real-world study of statins in people with HF found a protective effect of statins in heart failure in a large observational study. Subsequently, two RCTs of stains in heart failure found no benefit. In that case, we rely on the RCT.

Prof. Kosiborod: If you have RWE before an RCT, it should be regarded as hypothesis-generating only. We have a different situation here, since we already know what treatment effects are in RCTs, but we’re really trying to understand whether in clinical practice, where these medicines are used in very different way, we’re really trying to understand whether results overall are qualitatively similar. I wouldn’t obsess about point estimates. We shouldn’t expect effect sizes and hazard ratios to be the same, because they use completely different methodologies. It’s really important that they’re qualitatively similar, in the same direction. It tells a story that tells you what you should and shouldn’t be doing in clinical practice. CVD REAL fills gaps and completes stories that CVOTs didn’t tell us.

Q: Guidelines recommend choosing either liraglutide or SGLT-2 inhibitor for patients with existing CVD. On what basis do you prefer one molecule or another and in what indication, and specifically, along the lines of CVD, angina, stroke – in essence, which for which? Thanks.

Prof. Khunti: We’ve seen this in RCTs. CANVAS and EMPA-REG have certain consistent outcomes, in MACE, but stroke in EMPA-REG went in the wrong direction. In RWE, they’re mostly going in the right direction in terms of benefits. CVD REAL assessed SGLT-2 as a class, not looking to see if one is better than another. Guidelines, as I understand, are saying that therapies have proven benefits – empagliflozin, canagliflozin, and liraglutide – but not determining one is better than another. If you look at adverse events profile, there are differences. DECLARE will tell us if that is a class effect or if there is a difference between molecules.

Prof. Kosiborod: So you’re basically asking: A patient comes into clinic with type 2 diabetes, maybe had an MI five years ago, maybe a stroke, has symptoms of shortness of breath, may be at risk of heart failure, MI…which do I go with first? GLP-1 receptor agonist or SGLT-2 inhibitor? If I were to be an extremely data-driven person, I can’t answer your question based on clinical trial data. We also don’t have real world data that would guide decisions on which drug for which patient or which drug to use first. That becomes very difficult methodologically to analyze. The best I can tell you – that kind of question may never be answered in a clinical trial – is both agents are beneficial in patients like that. We don’t know mechanisms of action, GLP-1 and SGLT-2, but it’s becoming clear that it has little to do with glucose-lowering. It’s becoming clear that mechanisms are likely completely different. HARMONY showed a very robust reduction in MACE, DECLARE coming soon, positive on CV death and heart failure. GLP-1s tend to have potent effects on MACE/atherosclerotic endpoints, SGLT-2s have modest effects on MACE, but robust effect on things like heart failure. If we put cost aside, which of course isn’t realistic, but putting cost aside, there’s rationale for using both. It’s less important about which to use first, but important that we use everything in our armament. If cost is in the picture, we need to have a risk:benefit conversation with patient. That’s the best I can tell you with what we have today.

Q: RWE complements RCTs. Is that a regulatory requirement?

Prof. Kosiborod: Is RWE part of regulatory requirement? That depends on the regulator. What we do know in the US with FDA is there is now a mandate from the 21^st Century Cures Act to say that RWE needs to be incorporated in regulatory decision-making. How to do that is left up to the FDA. They’re working on a guidance document to be released in future, so we’ll better understand how they plan to do that. It’s impossible to answer if something seen in RWE-generation, a safety signal, would lead to withdrawal. It all depends on the details: what the drug is, what the signal is, how common and dangerous the signal is, whether the signal is confirmed in multiples studies. It’s not enough to find a signal in one database, you need to make sure it’s real. You may need to do additional clinical trials to see if the signal is consistent across clinical settings.

Oral Presentations: SGLT2 Inhibitors: New Mechanisms and Clinical Evidence

Empagliflozin Reduces Mortality and Hospitalisation for Heart Failure Irrespective of Cardiovascular Risk Score at Baseline

David Fitchett, MD (University of Toronto, Canada)

Dr. David Fitchett presented results from an intriguing EMPA-REG OUTCOME post hoc analysis showing that empagliflozin’s effect on reducing mortality and hospitalization for heart failure applied across a spectrum of baseline CV risk. To assess whether empagliflozin’s benefits varied across the CV risk continuum, patients in EMPA-REG OUTCOME (both placebo and treatment arms) were stratified into levels of low, intermediate, high, and highest CV risk based on TIMI Risk Scores for Secondary Prevention (TRS 2^oP). For context, the 10-point TRS 2^oP score includes 10 characteristics associated with CV risk (one point for each characteristic). Approximately 12% of study patients were binned into the low CV risk group, with ~40% in the intermediate risk group, ~30% in high risk, and the remaining ~18% in highest risk group. As expected, the percentage of patients experiencing an adverse CV event over the course of the CVOT increased as the level of risk elevated between the subgroups (figure below). Most interestingly, looking at various outcomes (including CV death, all-cause death, 3-point MACE, hospitalization for heart failure, and hospitalization for heart failure or CV death), benefit consistently trended towards empagliflozin across all levels of risk subgroups (see charts below), with interaction p values all at p=~0.8 or above –reflecting very low heterogeneity between risk subgroups. Dr. Fitchett therefore commented that these findings suggest the benefits of empagliflozin treatment extend across the spectrum of those with both type 2 diabetes and CVD, regardless of baseline CV risk. Although all patients in the EMPA Reg Outcome study had established CVD, the spectrum of risk in this population is very wide. Furthermore, the event rates in the lower risk group are similar to that observed in the patients without established CVD in the CANVAS study.    The DECLARE CVOT with dapagliflozin showed a reduction of HF but not CV death in the population without CVD.

SGLT2i vs Bolus Insulin as Add-on to Stable Basal Insulin Treatment in Type 2 Diabetes and Risk of Cardiovascular Disease and Mortality: A Nationwide Observational Study

Jan Erisson, MD, PhD (Uppsala University, Sweden)

Dr. Jan Erisson presented the results of a nationwide observational study in Sweden (n=2,988), which associated SGLT-2 inhibitors with a significantly reduced risk of hospitalization for heart failure or CV mortality (composite HR=0.51, 95% CI: 0.32-0.82, p=0.005), three-point MACE (HR=0.54, 95% CI: 0.34-0.87, p=0.011), severe hypoglycemia (HR=0.61, 95% CI: 0.40-0.93, p=0.022), and all-cause mortality (HR=0.30, 95% CI: 0.18-0.51, p<0.001) compared to bolus insulin, both as an add-on to stable basal insulin treatment in patients with type 2 diabetes. The study, which employed propensity score matching, followed Swedish adults with ≥3 dispenses of basal insulin within the last 12 months (between 2013 and 2016) who started either an SGLT-2 inhibitor or bolus insulin (n=1,494 each) for an average of 0.92 years. Notably, dapagliflozin accounted for ~79% of SGLT-2 usage, followed by empagliflozin at ~21%, and canagliflozin with <1%. Bolus insulin usage was 1% human, 56% insulin aspart, 28% insulin lispro, and 15% insulin glulisine. The significant improvements in HHF or CV mortality, MACE, and severe hypoglycemia held for many sensitivity analyses, though the 95% CI crossed unity when other definitions of basal insulin were allowed (see below). This study reinforces already-strong RCT and real-world evidence for cardioprotection as a class effect of SGLT-2 inhibitors, and it also helps to demonstrate that these benefits extend to a wide range of patients – only ~32% of participants in each group had baseline CV disease. To be sure, this is a great example of how real-world data can fill gaps in knowledge and help paint a more complete picture of novel agents and treatment algorithms.

Oral Presentations: Foot Prints

Association of Diuretics Use and Amputations in Patients with Type 2 Diabetes: A Hypothesis Driven from CANVAS Warning?

Ronan Roussel, MD, PhD (Centre de Recherche des Cordeliers, Paris, France)

Dr. Ronan Roussel presented a long-term (median follow-up 7.2 years, until December 2015) observational study of the SURDIAGENE French cohort (n=1,459 participants with type 2 diabetes from a single center), demonstrating that diuretic use was independently associated with risk of lower limb amputations (adjusted HR=2.03, 95% CI: 1.19-3.46, p=0.013). Importantly, these results give further credence to the hypothesis that the amputation signal found with J&J’s SGLT-2 inhibitor Invokana (canagliflozin) in CANVAS was driven by the molecule’s diuretic effect, which leads to decreased peripheral perfusion. The study also examined a propensity matched cohort (n=1,074), in which the association for lower limb amputations remained significant (adjusted cox proportional HR=2.13, 95% CI: 1.17-3.87). Notably, the hazard ratios for the primary composite outcome of first occurrence of a lower limb event (amputation or revascularization) in the full and matched cohorts were significant as well, though both were driven solely by amputations (see tables below). Dr. Roussel concluded by acknowledging the study’s weaknesses: (i) as an observational study, these results are only hypothesis-generating, not conclusive; (ii) propensity score matching is good, but not perfect; (iii) no data was available on the index date for drug initiation; (iv) the SURDIAGENE French cohort was a single-center cohort; and (v) residual confounding from measured and unmeasured confounders.

Cox proportional hazard ratio in propensity matched cohort (n=1,074, 536 diuretic users, 536 non-diuretic users)

	Lower Limb Events		Lower Limb Amputations		Lower Limb Revascularizations
	HR (95% CI)	p-value	HR (95% CI)	p-value	HR (95% CI)	p-value
Unadjusted	1.75 (1.19-2.58)	0.005	2.34 (1.35-4.05)	0.003	1.28 (0.82-2.01)	0.28
Adjusted for additional unbalanced covariates*	1.60 (1.06-2.42)	0.027	2.13 (1.17-3.87)	0.013	1.12 (0.70-1.79)	0.6443

*Covariates were baseline hypertension, RAAS blockers, beta-blockers, and statins use

Sensitivity analysis in the full cohort (n=1,459, 670 diuretic users, 789 non-diuretic users)

	Lower Limb Events		Lower Limb Amputations		Lower Limb Revascularizations
	HR (95% CI)	p-value	HR (95% CI)	p-value	HR (95% CI)	p-value
Crude	2.08 (1.49-2.93	<0.0001	2.53 (1.61-4.10)	<0.0001	1.71 (1.15-2.56)	0.01
Propensity score adjusted	1.74 (1.20-2.52)	0.003	2.20 (1.30-3.71)	0.003	1.31 (0.85-2.01)	0.23
Adjusted for cofounders*	1.72 (1.19-2.49)	0.0042	2.03 (1.19-3.46)	0.009	1.37 (0.89-2.12)	0.15

*Adjusted for age, diabetes duration, sex, BMI, smoking, systolic blood pressure, diastolic blood pressure, history of CV disease, history of LLE, urine albumin-to-creatinine ratio (UACR), eGFR, beta-blockers, lipid lowering drugs, and insulin use.

Oral Presentations: GLP1 Receptor Agonists, SGLT2 Inhibitors and the Kidney: New Lessons from Large Clinical Trials

Canagliflozin and Cardiovascular Outcomes in Patients with Chronic Kidney Disease

Vlado Perkovic, PhD (The George Institute for Global Health, Sydney, Australia)

A new CANVAS subgroup analysis revealed that canagliflozin’s CV benefits persist across different levels of renal function in study participants with renal impairment. In CANVAS, ~20% of enrolled participants had renal impairment (n=2,039), defined as eGFR <60 ml/min/1.73 m². As is known, in the overall population, canagliflozin conferred a significant 14% relative risk reduction for MACE events (HR=0.86, 95% CI: 0.75-0.97). There was no significant variation on this endpoint according to the degree of renal impairment among participants with eGFR <45, 45-<60, 60-<90, and ≥90 mL/min/1.73 m² (p=0.33 for interaction). The same pattern held true for the individual CV outcomes of CV death (overall HR=0.87, 95% CI: 0.72-1.06; p=0.53 for interaction), fatal and non-fatal MI (HR=0.89, 95% CI: 0.73-1.09; p=0.08 for interaction), and hospitalization for heart failure (HR=0.67, 95% CI: 0.52-0.87; p=0.62 for interaction) – see below for full results. Interestingly, fatal and non-fatal stroke did see a significant interaction based on renal function; the HR of 0.87 (95% CI: 0.69-1.09) for the overall population grew progressively weaker with higher eGFR (from HR=0.32 for eGFR <45 to HR=1.42 for eGFR ≥90; p=0.01 for interaction). However, the confidence intervals are quite wide, and the event rate was relatively lower than most other endpoints. Presenter Dr. Vlado Perkovic did not speculate on what might cause this effect on the stroke endpoint, and we remain intrigued as to whether this signal is real or a statistical fluke. Separately, it’s important to note that canagliflozin was less efficacious in people with an eGFR <60 vs ≥60 ml/min/1.73 m²  for both A1c (-0.43% vs. -0.64%, p<0.0001) and body weight (-1.16 kg vs. - 1.43 kg, p=0.0002). These results reinforce that canagliflozin’s effects on A1c and weight are conferred through a mechanism separate from the drug’s cardioprotective mechanism (of course, this is hardly new information!). In the end, this sub-analysis offers an important confirmation that, despite less impressive A1c reductions with canagliflozin in people with renal impairment, there is no attenuation of the SGLT-2 inhibitor’s cardioprotective effects in this population – supporting their use for cardiovascular risk reduction regardless of renal function (more on this below). This is critically important given the especially elevated CV risk that comes with comorbid diabetes and CKD. Though not mentioned in this oral presentation, it’s worth noting that canagliflozin also (i) demonstrated a likely renal protective effect in CANVAS; and (ii) has apparently demonstrated definite renoprotection in a dedicated CKD trial, CREDENCE, which was recently stopped ~one year ahead of schedule due to overwhelmingly positive results. Other SGLT-2’s Farxiga (dapagliflozin) and Jardiance (empagliflozin) also have their own dedicated renal outcomes trials (Dapa-CKD, expected to complete in November 2020, and EMPA-KIDNEY, expected to begin by the end of 2018).

• These findings also have implications for the current recommendation that SGLT-2 inhibitors be discontinued in patients with an eGFR <45 ml/min/1.73 m² (in the case of Invokana and Jardiance) and <60 ml/min/1.73 m² (in the case of Farxiga and Steglatro). According to Dr. Perkovic, this sub-analysis adds to the growing body of evidence prompting the field to reconsider the use of SGLT-2 inhibitor therapy in people with severe renal impairment; not only do SGLT-2s pose little harm to this group, but they could also potentially play an active role in protecting the kidneys from further damage.

Oral Presentations: Clinical Use of Insulin: What Works and What Doesn’t

Comparison of Canagliflozin and Liraglutide as a Replacement for Bolus Insulin in Type 2 Diabetes Patients Well-controlled by Basal-bolus Insulin

Naoki Kumashiro, MD (Toho University, Tokyo, Japan)

Dr. Naoki Kumashiro took the stage to present results from a small (n=40) comparative study demonstrating the efficacy of canagliflozin (J&J’s Invokana) and liraglutide (Novo Nordisk’s Victoza) as replacement for bolus insulin in patients who were well controlled with basal/bolus regimens. Forty patients with well-controlled type 2 diabetes on basal/bolus therapy (baseline A1c 6.7%) were randomized to replacing their bolus insulin with either canagliflozin (fixed at 100 mg/day) or liraglutide (increasing from 0.3 to 0.9 mg/day; doses in Japan are lower). After 24 weeks, A1c was adequately maintained compared to baseline: In the canagliflozin group, A1c dropped a nominal 0.1% from 6.8% to 6.7%; in the liraglutide group, A1c fell 0.2% from 6.4% to 6.2%. Rates of overall hypoglycemia were not different between groups, while rates of nocturnal hypoglycemia decreased in both groups when compared to baseline (exact numerical data was not presented here). No severe hypoglycemia events occurred in either group, which is certainly positive. We were pleased to see that researchers used CGM in this study to track glucose fluctuations, though the data indicated that these fluctuations did not meaningfully change in either group from baseline; we would love to see time-in-range comparisons between the groups and compared to baseline measurements. As expected with canagliflozin and liraglutide, body weight and BMI also tended to decrease in both groups when compared to baseline; however, exact numbers were not presented here. Dr. Kumashiro also noted that the total score for diabetes-treatment-related quality of life (DTR-QOL) significantly improved with both groups when compared to baseline; in our view, the lower treatment burden of an SGLT-2 inhibitor or GLP-1 agonist over bolus insulin is where much of the benefit of this switch is experienced by patients. However, while all domain scores improved in the canagliflozin group, no improvement was see in the liraglutide group in the domains of anxiety and dissatisfaction with treatment; we wonder if this can be attributed to the injection burden or even GI side effects. Nevertheless, these data add to a growing body of evidence supporting the effectiveness of SGLT-2 inhibitors and GLP-1s as a replacement for bolus insulin in controlling post-prandial glucose levels, and our sense is that clinical practice is reflecting these improvements more and more. 

Corporate Symposium: Illuminating the Path of Care for a Multifaceted Disease (Sponsored by Sanofi)

One Drug, Two Mechanisms: An SGLT-1 and SGLT-2 Inhibitor

Julio Rosenstock, MD (University of Texas, Dallas, TX)

Dr. Julio Rosenstock offered a glimpse into the possible future of SGLT outcomes studies: “If there was ever a need for a pragmatic CVOT, this is it – SGLT inhibitors in people with type 1 diabetes at high risk for CV disease.” He reminded the audience that elevated CV risk with diabetes is not unique to type 2, and noted that “if we’re all so excited about SGLTs for CV disease, heart failure, and nephropathy in type 2 diabetes, we need to honestly ask ourselves if these effects can be replicated in type 1.” At ADA 2018, Dr. David Cherney offered a similar call for a cardio/renal outcomes trial in type 1 diabetes, and Dr. Anne Peters shared that many of her type 1 patients ask for an off-label SGLT-2 inhibitor primary for the cardioprotective benefits – this is clearly a topic on people’s minds, from patients to providers to KOLs. To be sure, we currently have zero hard evidence on organ protection with these agents in type 1, though the mechanisms of SGLT inhibition – which are not thought to be glucose-mediated – should theoretically play out in type 1 diabetes much like they do in type 2. While the financial and logistical challenges of a full-fledged CVOT in a type 1 population are many (though negligible compared to cardiovascular and renal costs), Dr. Rosenstock’s suggestion of a pragmatic trial design (whether funded by pharma or philanthropy) is an intriguing one – especially coming from one of the most respected (if controversial) clinical trialists in the field. We do think that ongoing clinical trials of SGLT-2s in both heart failure and CKD outside of people with diabetes could shed valuable light on the broad applicability of these cardio- and renal-protective effects, and we hope that some in these trials have type 1 – to our knowledge, Lilly/BI’s EMPEROR and EMPA-KIDNEY trials are enrolling type 1s, but type 1s are excluded from AZ’s Dapa-HF and Dapa-CKD. Given the enormous gaps in treatment for people with type 1, we hope to see more discussion and characterization of a CVOT in people with type 1 as well as, overall, more on the risk/benefit profile of SGLTs in type 1. 

Beyond Glucose Control: Living with Diabetes

Thomas Danne, MD (Hannover Medical School, Germany)

Dr. Thomas Danne focused on the role SGLT inhibitors can play in easing the emotional burden of type 1 diabetes management. Citing data from a Clinical Diabetes paper authored by Close Concerns, The diaTribe Foundation, and dQ&A, Dr. Danne framed time-in-range as perhaps the most important player in the emotional aspects of type 1 diabetes management, and he positioned SGLTs as a key strategy for improving time-in-range. According to a dQ&A survey of nearly 7,000 people with diabetes, the top three ways in which diabetes impacts day-to-day life for people with type 1 all relate to the constant vigilance required to achieve in-range glucose numbers. These factors were: (1) The time commitment and burden good diabetes management takes; (2) Difficulty managing blood glucose; and (3) Hypoglycemia. With potent improvements in time-in-range (pooled CGM data from Lexicon’s inTandem1 and inTandem2 trials showed an additional 1-3 hours spent in-range with Sanofi-partnered SGLT-1/2 inhibitor Zynquista and AZ’s DEPICT 1 study showed an additional 2-3 hours in-range with SGLT-2 inhibitor Farxiga), SGLT inhibitors could play a key role in offsetting the stress, worry, and constant effort required to achieve time-in-range with insulin therapy alone.

• We additionally appreciated Dr. Danne’s unique take on the DKA risk associated with SGLT inhibitor therapy in people with type 1 diabetes: “DKA has been so forgotten and undermanaged in our clinics that I personally think it’s good that SGLT inhibitors have put a spotlight on it.” To be sure, DKA risk is a legitimate concern, especially in light of DEPICT-2 results that revealed increased DKA incidence with AZ’s SGLT-2 inhibitor Farxiga (not seen in DEPICT-1), essentially solidifying DKA risk as a class effect for SGLTs in T1D at large. That said, we love the idea of reframing this as an opportunity for HCP and patient education on DKA prevention, risk management, and treatment, rather than a deal-breaking safety signal for an otherwise extremely promising drug class. We agree with Dr. Danne that this complication has been overlooked and under-treated for too long, and the silver lining here is that more attention is being placed on it (in a similar way, we see the Invokana amputation signal as an opportunity to enhance patient education on foot care in diabetes). Dr. Danne highlighted the ATTD-sponsored consensus meeting on DKA risk mitigation held at ADA 2018, hinting that a forthcoming manuscript will dive into the specifics of patient selection, education, and monitoring, as well as offer guidance on which patients are ideal candidates for SGLT inhibitor therapy, how often they should check ketones, and how – all extremely important issues with Ad Comm(s), as regulatory decisions on SGLT inhibitors in type 1 diabetes loom large. Get a full take on these questions here.

Corporate Symposium: Tomorrow’s Management of Diabetes, Today (Sponsored by AZ)

Keynote: Modern Management of Type 2 Diabetes – Glucose Lowering is Not Enough

John Buse, MD (University of North Carolina, Chapel Hill, NC)

In his keynote at AZ’s daylong Monday symposium, Dr. John Buse unequivocally advocated for SGLT-1/2 and SGLT-2 inhibitors to be approved and used as adjunct therapies for patients with type 1 diabetes. Dr. Buse referenced both the DEPICT (dapagliflozin in type 1) program, which demonstrated significant reductions in A1c, total daily insulin dose, and body weight with dapagliflozin vs. placebo, as well as the inTandem (sotagliflozin in type 1) program, which found superiority of sotagliflozin over placebo on a plethora of endpoints. He briefly touched on the EASE program (empagliflozin in type 1; topline reported at ADA 2018), highlighting that the full results presentation on Thursday is what he is most excited for at EASD. On what he called the “fly in the ointment” – DKA, particularly euglycemic DKA – Dr. Buse stressed that both industry and clinicians are working diligently to ensure that SGLT inhibitors are used safely in type 1 diabetes. In his words: “I do believe the benefits for patients with type 1 diabetes are important. Now we need to find safe ways to use them.” We could hardly agree more, but we also believe that many critical specifics on precisely how to use them safely need to be worked out, and implementing a robust education and safety program across the multiple stakeholders at play is no small task. For more answers on this front, we look very forward to the forthcoming manuscript spearheaded by Dr. Danne and ATTD (see above).

Corporate Symposium: SGLT-2 Inhibitor – Why Not Use One? (Sponsored by AZ)

Panel

Richard Holt, PhD (University of Southampton, UK), Mikhail Kosiborod, MD (University of Missouri, Kansas City, MO), Juris Meier, MD (St. Josef Hospital, Bochum, Germany), Jiten Vora, MD (University of Liverpool, UK)

In AZ’s Wednesday night mini-symposium “SGLT-2 Inhibitors – Why Not Use One?” we heard from a superstar panel of leading cardiologists and diabetologists. Drs. Mikhail Kosiborod, Juris Meier, Jiten Vora, and Prof. Richard Holt unsurprisingly thrust their collective support behind SGLT-2s, touching on the widely-debated safety profile of the class, benefits beyond glucose lowering, and the highly-anticipated DECLARE CVOT for AZ’s Farxiga (dapagliflozin). Here were some of our favorite moments:

On SGLT-2 Safety and Dapagliflozin:

• “There is an amputation signal with canagliflozin, which is a rather non-selective inhibitor of not only SGLT-2, but also SGLT-1 to a certain extent. However, this has not occurred in the phase 3 trials for more selective inhibitors of only SGLT-2, including empagliflozin and dapagliflozin […] The same is seen for low trauma fractures, suggesting that both might be a function of non-selective SGLT 1/2 dual inhibition.” – Dr. Juris Meier (editor’s note – we are also curious about the impact of trial design)

• “As it stands, there is no data suggesting that amputations are a class effect of SGLT-2s […] Both the empagliflozin and dapagliflozin development programs have not demonstrated a signal for amputation.” – Dr. Mikhail Kosiborod (editor’s note – there was a signal on the “real-world” trial shown at ESC ….)

• “Fournier’s gangrene is not that surprising for a drug with increased genitourinary tract infections (GTIs). Even with the high numbers of patients treated with SGLT-2 inhibitors across the US, there were only 12 [documented] cases. It is something we need to keep in mind, but they are very rare.” – Dr. Meier

On Benefits Beyond Glucose Lowering:

• “Ladies and gentlemen, what do I want from a diabetes treatment in 2018? Improvement in blood glucose, improvement in quality of life, and reductions in long-term micro- and macrovascular complications. With the SGLT-2 class, we are starting to see ways in which these will come together in one therapy.” – Prof. Richard Holt

• “What I think is the unexpected jewel in the crown of SGLT-2 inhibitors is the emerging idea that they are also beneficial for the kidney.” – Dr. Holt

• “What is becoming pretty obvious to us in the CV field is that heart failure is not only the most common comorbidity of diabetes but also the one that is associated with the worst prognosis […] SGLT-2s are the first class of drugs for diabetes with a very promising signal in prevention of heart failure.” – Dr. Kosiborod

• On dapagliflozin not achieving superiority on 3-point MACE in DECLARE: “We need to look at the data. One thing I will remind you of is that p-values are not end-all, be-all’s. They are highly dependent on the patient population. When you have agents that are less potent on MACE, such as SGLT-2s, and a high primary prevention population, it makes it a lot less likely from a statistical perspective that superiority in MACE will be found.” – Dr. Kosiborod

• “[Using SGLTs for type 1 diabetes] are good idea but not licensed yet – keep an eye on the data […] It is a situation in which you will need to keep an eye on DKA, but all studies presented here suggest great efficacy and that patients love them.” – Dr. Jiten Vora

Additional Comments

• Interestingly, session chair Dr. Danilo Verge (VP for Cardiovascular and Metabolic Disease, AZ) made a concerted effort to mention AZ’s Epanova (omega-3-carboxylic acids) as a “future therapy” for CV disease, citing recent excitement from the REDUCE-IT CVOT for Amarin’s Vascepa as sparking interest in the area of prescription-strength omega-3s for cardioprotection. Epanova is under investigation in the STRENGTH CVOT, expected to complete October 2019.

•  “I would posit to you that, as far as I am concerned, the goal of treating diabetes should be to prevent morbid complications of diabetes.” – Dr. Kosiborod

• “The simple fact that CV disease kills most patients with diabetes is not as well known in the clinic as it should be.” – Dr. Kosiborod

• “More than half of people with diabetes worldwide do not undergo treatment intensification until their A1c is greater than 8.0%.” – Dr. Holt

• “Only 10% of patients who could benefit from an SGLT-2 are taking one.” – Dr. Vora

Posters

The inTandem1 Study: 52-week Efficacy and Safety of Sotagliflozin, a Dual SGLT1 and SGLT2 Inhibitor, as Adjunct Therapy to Insulin in Adults with Type 1 Diabetes

SK Garg, J Buse, J Rosenstock, T Bailey, P Banks, BW Bode, T Danne, JA Kushner, WS Lane, P Lapuerta, DK McGuire, A Peters, J Reed, S Sawhney, and P Strumph

A fascinating Lexicon poster illuminated the patient perspective on SGLT-1/2 inhibitor therapy, thorough the lens of participants in the inTandem1 and inTandem2 trials. 75 of the 1,575 study participants were contacted for a post-study phone interview, and 41 accepted. On type 1 diabetes in general, these participants reported a general lack of blood sugar stability (n=35), frequent highs (n=40) and lows (n=32), higher-than-desired A1c levels (n=37), and increased insulin use over time (n=24). They additionally reported a notable impact of type 1 diabetes on their day-to-day life, in terms of physical well-being (n=34), emotional well-being (n=31), impact on daily life (n=27), productivity and physical activity (n=20), feelings of stress and worry (n=21), feeling depressed or irritable (n=21), and overall quality of life (n=19). Against this rather bleak backdrop, interviewees in the active treatment arm of inTandem1 (n=27) saw a treatment satisfaction rise from a mean of 3.9 to 5.5 (on a 0-6 scale) before and after the study, vs. a decrease in treatment satisfaction over the course of the study (from 4.3 to 3.7) for those on placebo (n=14). This ringing endorsement for sotagliflozin therapy was even more apparent at a more granular level: Over the course of the study 81% of sotagliflozin-treated participants noted fewer high blood sugar events and greater blood sugar stability (vs. 36% of interviewees on placebo), 48% reported that insulin seems to work better (vs. 21%), and 37% reported fewer low blood sugar events (vs. 14%). Also to this point, a tremendous 89% of participants on active sotagliflozin treatment (n=24) stated that they would have liked to remain on the study medication, vs. only 43% of interviewees on placebo (n=6). We’re thrilled to finally have this concrete evidence to back up what type 1 patients and their physicians have been saying about SGLTs for type 1 all along: You can truly feel the time-in-range difference with an SGLT inhibitor, and it makes diabetes management less burdensome. We’re equally thrilled to see a large clinical trial evaluating a therapy’s emotional and real-life impact in addition to its clinical impact. We do, however, note that it’s very possible those who agreed to the phone interview could have been disproportionately pleased with sotagliflozin treatment.

• We were particularly struck by the direct patient testimony Lexicon included in the poster:

  • “The stabilizing of [blood sugar] gave me more energy and less worry. Being able I guess to get my normal life back – as normal as it can be with diabetes.”

  • “It was narrowing the band. Not having those contrasting highs and lows as frequently was the big improvement.”

  • “[Before] my sugar would keep bouncing around no matter what I did. It would go high; it would go low. It was not stable.”

• Very notably, patients appeared not to be overburdened by inTandem1’s ketone monitoring requirement. Nearly half of the interviewees (n=20) reported undergoing at least occasional ketone monitoring both before and during the trial, and among this group, all reported believing that they could avoid DKA by both monitoring ketones and understanding the signs and symptoms of hyperglycemia. Although to date there is no exact consensus on how to manage DKA risk in people with type 1 diabetes on SGLT inhibitor therapy, future treatment recommendations will undoubtedly involve ketone monitoring in some form, and it’s reassuring that this isn’t too arduous for patients to incorporate into daily life. That said, we would have liked to see a greater proportion of participants actively use ketone monitoring (our understanding is that trial protocol did not recommend regular checks).

Meet the Expert Sessions

MSD Meet-the-Expert

Sam Dagogo-Jack, MD (University of Tennessee Nashville, TN)

In an MSD Meet the Expert session, University of Tennessee Health Science Center’s Dr. Sam Dagogo-Jack emphasized that 22.4% of the cohort for VERTIS CV, the Steglatro (ertugliflozin) CVOT slated to complete in September 2019, has a history of heart failure. “Nearly one in four with a history of heart failure makes this study uniquely positioned to look at heart failure in a more robust fashion – it’s nearly twice what was enrolled in EMPA-REG OUTCOME.” Of course, EMPA-REG OUTCOME did show a 35% relative risk reduction in heart failure hospitalization; how will ertugliflozin stack up? To our understanding, VERTIS-CV is also the only SGLT-2 inhibitor CVOT to date to collect baseline heart failure status in all participants, making it uniquely well positioned to examine this effect. Click here for a deeper dive on heart failure and SGLT-2 inhibitors. More broadly, 99.9% of the cohort has had a prior CV event, firmly establishing VERTIS CV as a secondary prevention study.

Insulin Therapy

Oral Presentations: Where, When, and How of Rapid Acting Insulins

The Ultra-Rapid Insulin BioChaperone Lispro Shows Favorable Pharmacodynamics and Pharmacokinetics Compared to Faster Insulin Aspart and Insulin Aspart in Insulin Pumps

Gregory Meiffren, PhD (Adocia, Lyon, France)

A randomized controlled, double-blind, three-period crossover trial (n=43 type 1s) compared the pharmacodynamics and pharmacokinetics of BioChaperone Lispro, insulin aspart, and Fiasp (faster-acting insulin aspart) by pump delivery. While total insulin exposure was found to be similar between the three insulins, BioChaperone Lispro (AUC_{INS 0-1h} 68±27 h*mU/L; AUC_{INS 2-6h} 81±43 h*mU/L; t_{late0.5INSmax} 147±48 min) was associated with higher early insulin exposure, lower late exposure, and earlier time to late half-maximum exposure as compared to Fiasp (AUC_{INS 0-1h} 59±21 h*mU/L; AUC_{INS 2-6h} 93±45 h*mU/L; t_{late0.5INSmax} 165±59 min) and even more significantly to insulin aspart (AUC_{INS 0-1h} 43±22 h*mU/L; AUC_{INS 2-6h} 95±41 h*mU/L; t_{late0.5INSmax} 183±68 min). BioChaperone Lispro also had significantly faster-on and faster-off performance compared to insulin aspart, with higher area under the glucose infusion rate curves during the first two hours (AUC_{GIR 0-2h} 592±275 mg/kg vs. 500±244 mg/kg, p<0.0038), lower AUC_{GIR 2-6h} (784±402 mg/kg vs. 980±453 mg/kg, p=0.0015) and earlier times to early and late half-maximum GIR (t_{early0.5GIRmax} 44±22 min vs. 58±19 min, p<0.0001; t_{late0.5GIRmax} 210±68 min vs. 232±52 min, p-0.002). BioChaperone Lispro demonstrated similar early glucose-lowering effects and a significantly lower late glucose-lowering effect relative to Fiasp. The investigators concluded that when administered with an insulin pump, BioChaperone Lispro exhibits ultra-rapid pharmacokinetic and pharmacodynamic properties relative to insulin aspart and performed favorably as compared to Fiasp.

 

Hypoglycemia with Mealtime Fast-Acting Insulin Aspart vs. Insulin Aspart Across Two Large Type 1 Diabetes Trials

Christophe De Block, MD, PhD (Antwerp University Hospital, Belgium)

A pooled analysis of Novo Nordisk’s Onset 1 (n=1143) and Onset 8 (n=1025) trials comparing mealtime Fiasp (faster-acting insulin aspart) and NovoLog (insulin aspart) revealed that while overall rates of severe (as defined by the ADA) or blood glucose-confirmed (<56 mg/dl) hypoglycemia were similar between insulins, nocturnal hypoglycemia rates were lower with Fiasp. Fiasp demonstrated a statistically and clinically significant 16% relative reduction in the rate of hypoglycemia (ETR: 0.84, 95% CI: 0.72-0.98; p=0.02). A significant difference was not observed for diurnal hypoglycemia. Still, as Dr. Christophe De Block pointed out in his presentation of the results, nocturnal hypoglycemia in particular remains a major barrier to achieving glycemic targets, not to mention imposes on quality of life.

• Given that different basal insulins were used in the trials (Onset 1: insulin detemir; Onset 8: insulin degludec) Dr. De Block acknowledged that it is difficult to make direct comparisons between the studies. Still, the results suggest a lower hypoglycemia rate in Onset 8 vs. Onset 1, perhaps due to the different basal insulin used. Interestingly, while Fiasp was found to be non-inferior to NovoLog in change from baseline A1c in both trials, Onset 1 saw a statistically significantly greater A1c reduction with Fiasp.

Oral Presentations: The Bottom Line: What's the Best Basal Insulin?

The Clinical Benefits of IDegLira in DUAL VII were Achieved While Using a Simple Regimen with Fewer Injections and Dose Adjustments Compared with Basal-Bolus Therapy

Eden Miller, MD (Diabetes Nation, Bend, OR)

Presenting a new DUAL VII analysis, Dr. Eden Miller demonstrated 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. From a patient perspective (even a very heterogeneous group), this is not at all surprising. As a reminder, DUAL VII investigated Novo Nordisk’s Xultophy (insulin degludec/liraglutide fixed-ratio combination) vs. basal-bolus therapy with Lantus (insulin glargine) and NovoLog (insulin aspart). Results reported at ADA 2017 demonstrated Xultophy’s non-inferiority to basal-bolus on A1c lowering, but also indicated superiority on a number of secondary outcomes, including an impressive 89% reduction in severe or blood glucose-confirmed symptomatic hypoglycemia rate, not to mention ~2.1 lbs of weight loss vs. ~5.8 lbs of weight gain with basal-bolus. Dr. Eden Miller explained that, while the clinical benefits of Xultophy are well-known among the academic community, there are additional and equally meaningful benefits that Xultophy provides in terms of treatment complexity and convenience. In the DUAL VII study, the daily regimen for patients in the basal-bolus treatment group included the use of two pens, two to five daily injections, and two to five daily SMBG tests (one for each injection); we also know that, if a patient requires very high doses of basal insulin, number of daily injections can quickly climb even higher depending on which type of insulin/pen they use. Conversely, the Xultophy treatment arm only required one pen, one daily injection, and one SMBG test daily; we often write about the “added convenience” Xultophy offers patients over other injectable therapies, but putting it into these numbers reveals just how stark the contrast is. Considering the superior clinical profile Xultophy offers over basal-bolus regimens PLUS this tremendous improvement in patient experience, it’s hard to imagine many patients would choose the latter over the former – particularly considering the GI side effects caused by GLP-1s are smoothed out in the combination drug. Of course, any patient should be able to make an informed decision about their treatment regimen – any some may choose basal-bolus – but our feeling is that this often doesn’t happen, whether due to issues of cost, lack of information, or provider resistance to combination therapies. Dr. Eden also noted that the decrease in injection requirements with Xultophy translates to lower patient burden, as previous work has demonstrated a strong link between patients’ perception of treatment burden and the number of daily injection they must take. Patient burden was also especially high in the basal-bolus treatment group with respect to insulin dose adjustments during the study period: While both groups required a similar number of basal insulin adjustments on average (~17 total adjustments for both groups over 26 weeks), basal-bolus treatment required 200 bolus insulin adjustments, while Xultophy treatment – by its very nature – required zero such adjustments (see figure below). Burden was also far lower for the Xultophy treatment group in terms of total insulin dose: After 26 weeks, patients in the basal-bolus group took an average of ~84 units of total insulin per day, compared to only ~40 units for the Xultophy group (p<0.0001); basal dose alone was also higher in the basal-bolus group than the Xultophy group. All in all, this is amazing analysis though we are shocked by anyone who is surprised – insulin is one of the very hardest drugs in the world to dose, and it is incredibly variable day-to-day. Xultophy’s glycemic-dependent nature, as well as the stability of next-gen basal insulin and Victoza, are both hugely exciting.

CONFIRM: A Comparative Effectiveness Study of Insulin Degludec and Insulin Glargine 300 units/ml (glargine U300) in Insulin-naive Patients with Type 2 Diabetes

Joseph Tibaldi, MD (New York Presbyterian Hospital, Queens, NY)

Dr. Joseph Tibaldi presented results from the real-world CONFIRM study demonstrating insulin degludec’s (Novo Nordisk’s Tresiba) superiority to insulin glargine U300 (Sanofi’s Toujeo) on A1c lowering, hypoglycemia, and treatment discontinuation. In the study, 4,056 total patients were analyzed using US electronic health records and split into two matched cohorts of patients either taking insulin degludec or insulin glargine U300. After six months, degludec was associated with a larger reduction in A1c from baseline when compared to glargine U300 (-1.5% vs. -1.2%, p=0.029 for comparison). Patients taking insulin glargine U300 were 37% more likely to discontinue treatment vs. those on degludec (p<0.01). End of study insulin dose was also 9% lower for the degludec patients (p=0.038). Degludec treatment was also associated with a 30% lower rate of hypoglycemia compared to glargine U300 (HR=0.70, p=0.045) and a significantly lower change in the proportion of patients experiencing one or more episode of hypoglycemia (HR=0.64, p<0.01). These hypoglycemia data robustly align with evidence from other real-world studies of degludec (EU-TREAT) as well as clinical trials (SWITCH and DEVOTE) as illustrated in the comparison chart below; however, they stand in strong contrast to results from the Sanofi-sponsored BRIGHT study, which randomized patients to either Toujeo or Tresiba and favored Toujeo on some endpoints. In the end, we don’t think comparing these two basal insulins is the most fruitful conversation to have: Both offer tremendous improvements over earlier basal insulin options, but just over 20% of total US basal insulin prescriptions go to Tresiba and Toujeo combined. Far more patients could benefit from the improved stability both options offer, and there’s more than enough room for both of these highly impressive insulins to succeed on the market.

• Dr. Tibaldi underscored that CONFIRM and other real-world data will complete the evidence base for insulin degludec. He noted that degludec has had impressive results from RCTs including SWITCH and DEVOTE. However, it’s still important to determine the effects of degludec when there isn’t “hand-holding of patients” like there is in RCTs, and in situations where clinicians may not perfectly follow the labeled indications of the product. We’re glad to see this emphasis on complementary real-world studies of diabetes therapies, and applaud this approach of fleshing out a full clinical profile for a product based on a combination of study methodologies. To be sure, real-world studies carry risk of inherent confounding that cannot be controlled for, but we do think RWD has an increasingly important role to play in a rapidly evolving diabetes ecosystem.

Oral Presentations: Clinical Use of Insulin: What Works and What Doesn’t

Comparison of Canagliflozin and Liraglutide as a Replacement for Bolus Insulin in Type 2 Diabetes Patients Well-controlled by Basal-bolus Insulin

Naoki Kumashiro, MD (Toho University, Tokyo, Japan)

Dr. Naoki Kumashiro took the stage to present results from a small (n=40) comparative study demonstrating the efficacy of canagliflozin (J&J’s Invokana) and liraglutide (Novo Nordisk’s Victoza) as replacement for bolus insulin in patients who were well controlled with basal/bolus regimens. Forty patients with well-controlled type 2 diabetes on basal/bolus therapy (baseline A1c 6.7%) were randomized to replacing their bolus insulin with either canagliflozin (fixed at 100 mg/day) or liraglutide (increasing from 0.3 to 0.9 mg/day; doses in Japan are lower). After 24 weeks, A1c was adequately maintained compared to baseline: In the canagliflozin group, A1c dropped a nominal 0.1% from 6.8% to 6.7%; in the liraglutide group, A1c fell 0.2% from 6.4% to 6.2%. Rates of overall hypoglycemia were not different between groups, while rates of nocturnal hypoglycemia decreased in both groups when compared to baseline (exact numerical data was not presented here). No severe hypoglycemia events occurred in either group, which is certainly positive. We were pleased to see that researchers used CGM in this study to track glucose fluctuations, though the data indicated that these fluctuations did not meaningfully change in either group from baseline; we would love to see time-in-range comparisons between the groups and compared to baseline measurements. As expected with canagliflozin and liraglutide, body weight and BMI also tended to decrease in both groups when compared to baseline; however, exact numbers were not presented here. Dr. Kumashiro also noted that the total score for diabetes-treatment-related quality of life (DTR-QOL) significantly improved with both groups when compared to baseline; in our view, the lower treatment burden of an SGLT-2 inhibitor or GLP-1 agonist over bolus insulin is where much of the benefit of this switch is experienced by patients. However, while all domain scores improved in the canagliflozin group, no improvement was see in the liraglutide group in the domains of anxiety and dissatisfaction with treatment; we wonder if this can be attributed to the injection burden or even GI side effects. Nevertheless, these data add to a growing body of evidence supporting the effectiveness of SGLT-2 inhibitors and GLP-1s as a replacement for bolus insulin in controlling post-prandial glucose levels, and our sense is that clinical practice is reflecting these improvements more and more. 

Corporate Symposium: Illuminating the Path of Care for a Multifaceted Disease (Sponsored by Sanofi)

The BRIGHT Study: The First Large Head-to-Head Comparison of Second-Generation Basal Insulin Analogs

Alice Cheng, MD (University of Toronto, Canada)

Dr. Alice Cheng provided granularity on the Sanofi-sponsored BRIGHT study (n=929), the first head-to-head comparison of next-gen basal insulin analogs glargine U300 (Sanofi’s Toujeo) and degludec (Novo Nordisk’s Tresiba). Results were recently published in Diabetes Care. Dr. Cheng built on the topline readout from ADA, which found a statistically significant decrease in incidence of hypoglycemia <70 mg/dl (OR=0.74, 95% CI: 0.57-0.97, p=0.03) and hypoglycemia <54 mg/dl (OR=0.63, 95% CI: 0.40-0.99, p=0.044) with Toujeo vs. Tresiba during the first 12 weeks of the study – the “titration period.” However, there was no significant difference in the latter 12 weeks of the study – the “maintenance period.” The overt implication from these data was that Toujeo conferred a 26% risk reduction for hypoglycemia when compared to Tresiba, although the marginal p-values (0.03 and 0.044) and the fact that this benefit disappeared during BRIGHT’s maintenance phase left the field less than convinced that Toujeo is truly superior to Tresiba on hypoglycemia. That said, we were intrigued by the data Dr. Cheng shared at EASD: She highlighted a significantly higher fasting plasma glucose with Toujeo compared to Tresiba over the study’s 24 weeks (LS-mean difference of 0.43 mmol/L, or ~8 mg/dl; 95% CI: 0.15-0.70). However, the Toujeo group did start at a higher mean FPG than the Tresiba group, and it appears the difference was maintained throughout the trial more than anything. There was no significant difference in fasting SMPG nor A1c (see below). However, A1c equipoise at ~7.0% was achieved by 24 weeks, which might indicate that slightly less glucose variability was seen with Tresiba vs. Toujeo (we expect more glycemic data on Wednesday). Dr. Cheng also shared hypoglycemia incidence by time of day – an “eyeball statistic” (i.e., without rigorous statistical analyses – we’re guessing these would be underpowered) – suggesting that the hypoglycemia benefit of Toujeo during the titration period came between 6 AM and 2 PM. For reference, Toujeo and Tresiba were administered between 6 PM and 8 PM. As shown below, Toujeo carried a slightly higher incidence of hypoglycemia between 12 AM and 4 AM, but Tresiba was associated with more hypoglycemia in the late morning/early afternoon. Full glycemic data will be presented in an oral presentation this coming Wednesday, but see more below on important BRIGHT design aspects – and why they do not account for the hypoglycemia difference. To be sure, we haven’t heard much commentary on why this titration period result occurred, and our sense is that most thought leaders consider both these next-gen basals to be highly-preferable options over a first-generation analog or human insulin, or that they see a slight edge to Tresiba over Toujeo (after all, Novo Nordisk’s product now has hypoglycemia data showing a benefit vs. Lantus on its US and EU labels). We note as well that Novo Nordisk is conducting its own head-to-head trial of Tresiba vs. Toujeo, expected to complete in March 2019 per ClinicalTrials.gov.

 

• Starting doses for the two insulins differed due to labeling, though titration intensification followed the same schedule. As such, Dr. Cheng believes it unlikely that titration protocol contributed to the difference in hypoglycemia during titration. On the other hand, it seems possible to us that that same titration schedule may be more ideal for one insulin than the other; we know that Tresiba’s half-life is longer than Toujeo’s and wonder if this might be relevant. The BRIGHT study was conducted in North America and Europe, where labels indicate a starting dose of 0.2 U/kg for Toujeo and 10 U for Tresiba. Both were titrated weekly to a target fasting SMPG of 80-100 mg/dl without hypoglycemia (see below), and both therapies conferred a similar (~2 kg, or ~4.4 lbs) increase in body weight.

  • Notably, BRIGHT participants were required to have no prior insulin use and were allowed to continue any and all non-insulin anti-hyperglycemic agents during the trial. As such, Dr. Cheng dispelled the notion that the hypoglycemia difference between Toujeo and Tresiba could have been due to greater use of SU/glinide use in the Tresiba group. There was actually no statistically significant difference in hypoglycemia incidence between those taking an SU/glinide and those who were not, in either of the study groups, at either definition of hypoglycemia (p=0.578 for <70 mg/dL and p=0.733 for <54 mg/dL). There was also no difference in A1c reduction regardless of SU/glinide use.

• “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.” We couldn’t agree more with this praise from Dr. Cheng! As we see it, both products offer significant improvements over first-gen insulins (i.e. Sanofi’s Lantus and Novo Nordisk’s Levemir), in terms of glucose control and hypoglycemia. Moreover, BRIGHT offers strong evidence supporting the efficacy of basal insulin when truly titrated to target. That being said, Dr. Cheng did also use this evidence to suggest that the hypoglycemia difference between Toujeo and Tresiba was not due to a glycemic difference. However, we also find it notable that fasting plasma glucose trended significantly lower for Tresiba than Toujeo (see above), which Dr. Cheng did not address in depth.

  • In keeping with Dr. Bergenstal’s sentiments from the opening lecture of Sanofi’s symposium (see above), Dr. Cheng emphasized that the average starting A1c’s of both groups – 8.6% for the Tresiba group and 8.7% for the Toujeo group – were much too high. As she sees it, there should not be an A1c threshold at which insulin therapy is initiated in patients with type 2 diabetes. Rather, this decision should be governed by whether the patient can achieve his/her personal glucose targets without insulin; if not, insulin should be initiated regardless of A1c.

• Dr. Cheng noted that early hypoglycemic events (i.e. during titration) are associated with increased rates of treatment discontinuation, as well as long-term risk of hypoglycemia – of course, both Tresiba and Toujeo cause less hypoglycemia than first-gen basal insulins. On the former, Dr. Cheng referenced this retrospective cohort study, which found that 68% of patients who experienced hypoglycemia in the first six months following basal insulin initiation discontinued, compared to a 54% discontinuation rate in those who did not experience hypoglycemia (p<0.0001). On the latter, Dr. Cheng pointed to this retrospective longitudinal analysis of EMRs which demonstrated that hypoglycemia during the initial three month period of insulin therapy was associated with longer-term risk of these events over the ensuing 3-24 months (OR=5.71, 95% CI: 4.67-6.99) – an exceedingly significant association, though we imagine it’s also very likely that this relationship simply predicts/reflects baseline risk for hypoglycemia more than anything else. In the end, we again think the leap forward that the whole next-gen basal insulin class represents should be the takeaway.

Beyond RCTs: A Comprehensive Approach to Understanding Diabetes

Luigi Meneghini, MD (UT Southwestern, Dallas, TX)

UT Southwestern’s Dr. Luigi Meneghini provided updates on the DELIVER trial program, a series of real-world studies examining the effects of switching to Toujeo (Sanofi’s insulin glargine U300) from another basal insulin. In particular, Dr. Meneghini highlighted results from the new DELIVER D+ study, comparing glycemic outcomes in people with type 2 diabetes who switched from a first-gen basal insulin to either Toujeo or Tresiba. First presented in poster form at ADA 2018, this retrospective cohort study of EMR data found a significant reduction in overall hypoglycemia upon switching to Toujeo (from 15.6% to 12.7%, p=0.006), but only a non-significant numerical reduction in hypoglycemia upon switching to Tresiba (from 14.3% to 12.7%, p=0.12). While this is roughly consistent with the BRIGHT study data, it is also possible that Toujeo’s apparent superiority is an artifact of higher hypoglycemia rates in this cohort prior to insulin switching – that is, both insulins give the same exact final incidence of hypoglycemia. Is it possible that the population being switched to Toujeo, for whatever reason, was higher-risk at baseline? In our view, it’s critical not to miss the forest for the trees here; it’s clear that both next-gen basal insulins offer meaningful reductions in hypoglycemia. Dr. Meneghini further unveiled that topline results from the DELIVER Naïve D study point to similar A1c reductions and comparable rates of hypoglycemia with Toujeo and Tresiba in people with type 2 diabetes who are new to basal insulin therapy. He didn’t specify when we can expect the full readout; to our knowledge, exact numbers are not yet available. As a reminder, the DELIVER program also includes the DELIVER 2 and DELIVER 3 studies. DELIVER 2 showed that switching to Toujeo from another basal insulin conferred an estimated $1,439 of healthcare savings per patient per year, driven by lower hypoglycemia-related healthcare utilization – namely, hospitalizations, ER visits, and outpatient care. DELIVER 3 demonstrated that switching to Toujeo vs. other basal insulins was associated with significantly less hypoglycemia and similar glycemic control among older patients (≥65 years) with type 2 diabetes.

• In an ensuing panel discussion, an audience member made an astute observation: In real-world studies of next-gen insulins, how can we disentangle the effects of the next-gen agent from the higher quality of care that patients likely receive from HCPs who are savvy enough to prescribe these agents? Indeed, the early adopters of new therapies tend to be expert physicians who are delivering leading-edge care, and we imagine there could also be confounders in the patients they see (i.e. do experts more often treat patients with more resources and access to better care?). Of course, the irony here is that next-gen drugs tend to be safer and could therefore be used more broadly and easily by those who don’t specialize in treating diabetes, but in the real world we see precisely the opposite trend.

Novel Therapies

Symposium: Single and Combined Stimulation of GIP and GLP-1 Receptors in the Treatment of Type 2 Diabetes

Efficacy and Safety of LY3298176, a GIP/GLP-1 Receptor Co-agonist, in Subjects with Type 2 Diabetes: A Phase 2b Study

Juan Frias, MD (National Research Institute, Los Angeles, CA)

In a very highly anticipated phase 2 readout, Dr. Juan Frias revealed highly impressive glycemic and weight loss results for Lilly’s LY3298176 (“LY”), a once-weekly GIP/GLP-1 dual agonist  (see Lilly’s press release and a publication in The Lancet). In this phase 2b, 26-week dose-ranging study (n=315), four doses of LY (1 mg, 5 mg, 10 mg, and 15 mg) were examined in comparison to both Trulicity (dulaglutide) 1.5 mg and placebo – we’ve tabulated all of the efficacy and safety results below. In addition to those prespecified endpoints, two additional, exceptional results were also presented: Weight loss ≥15% and percent achieving A1c <5.7%. On the former weight loss endpoint, 22% of those on LY 10 mg and 25% of those on LY 15 mg achieved at least 15% weight loss, which was significant vs. both placebo and Trulicity. Additionally, 6% of those on LY 5 mg achieved this weight loss, but this result was not significant. Moreover, 18% of those on LY 10 mg and 30% on LY 15 mg achieved an A1c <5.7%, both significant against placebo and Trulicity. For comparison, 2% of the placebo group, 2% of the Trulicity group, and 4% of the 5 mg group achieved this result; these are ITT analyses, including all randomized participants. Of note, however, data presented below are on-treatment analyses, excluding those who discontinued study drug or initiated rescue therapy (the statistical principle Novo Nordisk has utilized in presenting phase 3 data for oral semaglutide). Still, all of these results reflect truly remarkable levels of efficacy, particularly at higher doses; the room truly came alive when Dr. Frias revealed the weight loss data in particular. To this end, we hope very much that Lilly will pursue LY for an obesity indication too, as there’s no doubt in our minds that more efficacious obesity pharmacotherapies that are accessible are desperately needed – so far, Novo Nordisk has really grown this industry from scratch with Saxenda. During Lilly’s 2Q18 update, management announced that LY would advance to phase 3 for type 2 by early 2019 – the first-ever dual agonist to make it to phase 3. Lilly also alluded to impressive efficacy, specifically by announcing that an investor call would take place following Dr. Frias’ presentation, and we have to admit: These results did not disappoint.

Efficacy Endpoints for LY3298176

	Study Arm
Endpoint (26 weeks; on-treatment)	LY 1 mg (n=52)	LY 5 mg (n=55)	LY 10 mg (n=51)	LY 15 mg (n=53)	Trulicity 1.5 mg (n=54)	Placebo (n=51)
A1c Reduction	0.7%*	1.6%*#	2.0%*#	2.4%*#	1.1%*	(0.1% increase)
Weight Loss	0.9 kg (2.0 lbs)	4.8 kg (10.6 lbs)*	8.7 kg (19.2 lbs) *#	11.3 kg (24.9 lbs) *#	2.7 kg (6.0 lbs)*	0.4 kg (0.9 lbs)
% Achieving  A1c ≤6.5%	15%*#	64%*#	82%*#	59%*#	39%*	2%
% Achieving A1c <7.0%	33%*	69%*#	99%*#	77%*#	52%*	12%
% Achieving ≥5% Weight Loss	14%	47%*#	71%*#	63%*#	22%*	0%
% Achieving ≥10% Weight Loss	6%	16%*	39%*#	38%*#	9%	0%

* = p<0.05 vs. placebo

# = p<0.05 vs. Trulicity

• As evidenced below, if there’s one concern over this candidate, it’s tolerability – and that’s a big concern. In our minds, these data (below) evoked an issue Sanofi’s recently saw with its GLP-1/glucagon dual agonist in phase 2 (the candidate is still advancing to phase 3 for obesity, as far as we know). However, one quarter of patients taking Sanofi’s SAR425899 dropped out due to tolerability issues – coincidentally equivalent to LY 15 mg. We think there’s a good chance that tolerability issues being more prevalent with dual agonists than standalone GLP-1 agonists might limit the uptake of these agents, particularly in lower-risk patients. Presumably, tolerability was also an issue with Trulicity to some extent. If, however, it is not too hard to work out which patients can tolerate the high dose easily – that’s where some magic would be.

Safety Endpoints for LY3298176

	Study Arm
Variable	LY 1 mg	LY 5 mg	LY 10 mg	LY 15 mg	Trulicity 1.5 mg	Placebo
Treatment Discontinuation	15%	15%	14%	34%	15%	18%
Tx Disc. Due to Adverse Events	4%	9%	6%	25%	11%	4%
GI Events	23%	33%	51%	66%	43%	10%
Nausea	4%	20%	22%	40%	30%	6%
Diarrhea	14%	24%	24%	32%	17%	4%
Vomiting	4%	7%	16%	26%	9%	2%
Total Hypoglycemia	2%	7%	10%	8%	4%	4%
Severe Hypoglycemia	0%	0%	0%	0%	0%	0%

• Dr. Frias noted that, for the 10 mg and 15 mg doses of LY, participants were titrated to those doses relatively quickly. In his assessment, this is likely responsible – in part – for the high prevalence of adverse GI effects; in an interview with our team, Lilly’s Dr. Brad Woodward (Global Brand Development Lead and Senior Medical Director) emphasized the same explanation. Indeed, GI side effects with 15 mg were concentrated in the earlier part of the trial but seemingly took a very long time to taper off (below). We’d add that these GI effects (nausea, vomiting, diarrhea) could also have augmented the weight loss observed, both by getting patients to the full dose more quickly (i.e. more time at most effective doses for weight loss), but also potentially through the GI side effects themselves. We expect more analyses on this front – we also think that might not be the case and overall we aren’t sure – this question always used to come up back in the day in phase 3 trials for Amylin’s Byetta. For the 10 mg dose, participants spent two weeks on 5 mg before escalating to 10 mg; for the 15 mg dose, participants spent two weeks at 5 mg then four weeks at 10 mg before escalating to 15 mg. Trulicity was started at 1.5 mg, and the 1 mg and 5 mg doses were not titrated.

  • We’re also eager to see any analyses aimed at teasing apart what mechanism drove glucose lowering. On this front, we’re curious what impact weight loss might have had on pushing down A1c (in addition to the well-characterized incretin effects of GLP-1) – an attendee also raised this issue during Q&A, noting that weight loss can unload the demand on the beta cell and actually result in a more durable long-term effect. Dr. Frias explained that insulin sensitivity did improve beyond what has been observed with selective GLP-1 agonists alone, so some improvement in glycemia is likely due to weight loss. We imagine the proportion of patients achieving an A1c <5.7% is much more meaningful if this effect is being driven by weight loss, rather than the artificial glucose-lowering mechanism of GLP-1 and GIP agonism.

  • We’ll be interested to see whether Lilly pursues the 15 mg dose in future trials. While the efficacy is certainly higher, that added benefit – in our view – is almost entirely offset by the lower tolerability. On the other hand, the 10 mg dose is about twice as efficacious on A1c lowering and three times as powerful on weight loss as Trulicity, with a much smaller jump in GI side effect incidence. If we had to pick, this is the one we’d zero in on. How will Lilly view these distinct profiles?

• Dr. Woodward couldn’t give us any insight on how Lilly might strive to demonstrate CV superiority with LY. It’s very clear at this point that CV benefit is a critical component of successful diabetes therapy today and hope that Lilly will consider a comprehensive investigation of LY’s impact on both cardiovascular and renal outcomes.

• At baseline, participants were what Dr. Frias characterized as a very typical population of people with type 2 diabetes. Across all six treatment arms, mean age was 50-56 years, diabetes duration 8-9 years, A1c 8.0%-8.2%, and BMI 32-33 kg/m². Given the impressive reductions on A1c and body weight, we might have expected a population with somewhat higher A1c and BMI – which makes the effects seen with LY all the more notable, in our view. On the other hand, this likely made it easier to achieve certain endpoints (i.e. A1c <5.7%). We find it interesting to think about this population, as we’d actually anticipate Lilly positioning this candidate for a set of patients farther along in the course of disease. Indeed, Dr. Frias’ first slide rightly noted that, despite highly-efficacious GLP-1 agonists coming to market in recent years, some patients still need greater efficacy on glucose-lowering and weight loss. Of course, this was a phase 2 dose-ranging study and Lilly will likely investigate LY in a variety of patient populations in phase 3.

• LY is a 39 amino-acid peptide based on the native GIP peptide, with modifications to allow GLP-1 receptor binding as well. Of note, the molecule actually has greater agonist activity than native GIP, but less than native GLP-1 – so it causes more GIP than GLP-1 activation. Mean half-life is five days, on par with once-weekly GLP-1s. Our observation is that no one truly yet understands the mechanism by which GIP offers additional benefit in the presence of GLP-1; we absolutely expect and look forward to increased scholarship in this area.

Commentary and Panel Discussion: Advantages of Combined Stimulation of GLP-1 and GIP Receptor in Type 2 Diabetes?

Matthias Tschop, MD (Helmholtz Zentrum München, Munich, Germany)

Following presentation of the very buzzworthy results just above, Dr. Matthias Tschop offered historical and scientific context on the results and Drs. Juan Frias and David D’Alessio joined him in responding to audience questions. Dr. Tschop explained that he started thinking about “twincretins” back in 2005, but it wasn’t always obvious to pair these two compounds up: There’s a lot of overlap in what they do, and many years were spent testing the hypothesis that we actually need GIP antagonists. However, there are important differences; for example, GIP seems to target adipose tissue more directly than GLP-1. Indeed, we also got the sense over the course of the meeting that balance between the two agonism components is very important, and GIP agonism alone doesn’t seem to be sufficient – Dr. Tschop raised the question, is there a therapeutic window for efficacy and tolerability in terms of balance? To complicate things further, GLP-1/glucagon dual agonists have their own unique features, including a dose-dependent decrease in body weight based on the level of glucagon agonism (which increases energy expenditure). Mechanistically, glucagon agonism would contribute to hyperglycemia, but these molecules don’t necessarily come with an increase in diabetes risk if they’re matched evenly with GLP-1 agonism – we find this feature and mechanism fairly un-intuitive. While the field wraps its head around these new mechanisms, it can find comfort in the fact that these molecules work: Impressive data from Lilly was joined by phase 2a data for AZ’s GLP-1/glucagon dual agonist MEDI0382. Both offer striking weight loss and improvements in glycemia; pending phase 3 data, dual agonists are poised to become the next big development in type 2 diabetes pharmacotherapy.

Q: I’m somewhat amazed by these results because GIP has been a bad guy in the past – if you knock it out, everything gets better. In human studies, there seems to be something different when they act together on cells. Is there any data about signaling in cells or ideas about why this changes GIP action so dramatically, especially since it’s primarily a GIP agonist?

Dr. D’Alessio: We have to be careful not to overstate what causes the great clinical effects. The simple model is that beta cells all have both receptors and these dual agonist molecules hit more of those receptors. Is that truly the model? Do all beta cells have both receptors or is there heterogeneity and we’re now hitting both sets? Is GIP activated or does the molecule actually act in humans as a super GLP-1 agonist? I showed mouse data where the dual agonist gives more weight loss but GIP by itself did nothing – we’ve known that for 40 years. When you look at molecules, we say they’re going to teach us a lot going backwards about physiology. I am not sure otherwise how to explain it, particularly the effect on weight.

Dr. Tschop: In summary, I’m equally unsure because we don’t have answers. One, we really failed to uncover over the last 15 years any cell-autonomous signaling interaction between these two that would fully explain these results. There could be a systemic interaction: improved glucose control through GLP-1 agonism that helps bring GIP-mediated benefits into action. Translational research could help – a global receptor knockout or pathway knockout may tell you, when you see metabolic effects, that there is an interesting pathway, but it rarely tells you which direction you need to go.

Dr. Francesco Giorgino: I have a question on the dose response you showed and the comparison with dulaglutide. If we look at data with the 5 mg dose, this was associated with less problems in reduced appetite but a more evident effect on weight loss and A1c. So can you speculate about mechanism or do you still need to look at feeding behavior in detail?

Dr. Frias: I think we would need more detail. With respect to the mechanism, we would expect that from a dual agonist at work with both GLP-1 and GIP components. We’re going to need more mechanistic work but I think that speaks to action from both GLP-1 and GIP agonism.

Dr. Steven Kahn: There’s clearly a beta cell effect from the incretin, but how much do you think will be from body weight loss and unloading the beta cell demand? In the long run there’s probably better hope for a durable effect if you can keep weight off.

Dr. Frias: I think yes. We looked at some insulin sensitivity measures and there is definitely improvement in insulin sensitivity beyond what we saw with selective GLP-1 agonism. Reductions in fasting insulin went along with that. We haven’t teased out how much is due to weight loss vs. other factors, but with this degree of weight loss at least some is probably due to weight loss.

Q: I was wondering if you looked at insulin secretion among those who got below an A1c of 5.7% – are you re-instituting GIP and insulin sensitivity, and did C-peptide levels change?

Dr. Frias: We’ve only seen fasting insulin so far.

Q: Can our impression about what GIP does or doesn’t do be limited by the fact that, until now, we’ve only seen very short term exposure to GIP? Maybe there is a time course involved that we need to know about.

Dr. Frias: It has been shown that some resistance will be alleviated by improvements in glycemic control. It’s possible that’s what we’re seeing here as well. Fasting glucagon was lower in patients  with the dual agonist.

Q: Have you looked at fat distribution?

Dr. Frias: There weren’t imaging procedures but you saw the waist circumstance data, so there is some hint at loss of visceral fat.

Q: Were there changes in cholesterol?

Dr. Frias: There was a very significant reduction in triglycerides. Cholesterol was not different from the relatively small reduction seen with GLP-1 agonists, but triglycerides were definitely further selectively reduced, possibly through GIP’s effect on adipose tissue.

Symposium: Combination Therapy for Type 2 Diabetes: Challenges and Controversies

Dual- and Tri- Agonists

Matthias Tschöp, MD (Helmholtz Zentrum München, Munich, Germany)

EASD 2018 was unquestionably a watershed conference for dual agonists, and wrapping up discussions on these emerging candidates – in one of Friday’s very last session – was Dr. Matthias Tschöp. He briefly reflected on the impressive new data presented by Lilly and AZ for GLP-1/GIP dual-agonist tirzepatide and GLP-1/glucagon dual-agonist MEDI0382, respectively, and discussed the remarkable potential of tri-agonists. On tirzepatide’s results, he cautioned against “comparing apples to oranges” with respect to Trulicity (dulaglutide), the active comparator in the study; notably, tirzepatide is a 39 amino-acid peptide based on the native GIP peptide, with modifications to allow GLP-1 receptor binding as well. As such, dulaglutide and tirzepatide are structurally (and mechanistically) very different, albeit both capable of GLP-1 receptor agonism. The remainder of Dr. Tschöp’s talk was dedicated to the future of multi-agonists. He emphasized that, with dual agonists, their demonstrated preclinical efficacy translated remarkably well to clinical trials. If this principle holds for tri-agonists in the pipeline, it would bode extremely well for this next-next-generation of incretins, which have demonstrated an additive effect on weight loss and glucose lowering in preclinical studies, compared even to dual agonists (though he acknowledged that it is way too early to be saying too much). In this excitement, Dr. Tschöp shared a list of all co- and tri-agonists currently in development (slightly updated from this recent paper) which includes 12 GLP-1/glucagon dual-agonists, five GLP-1/GIP dual-agonists, and three GLP-1/glucagon/GIP tri-agonists (we understand there are actually four, if one considers Zealand’s portfolio in addition to Novo Nordisk’s, Hanmi’s, and Sanofi’s molecules). To conclude his presentation, Dr. Tschöp offered idyllic projections for a future of “personalized precision medicines” for obesity and type 2 diabetes. As he sees it, a variety of dual- and tri- agonists have the potential to allow greater responsiveness to medication based on better individualization to patient needs and medical history, and we’d also love to see precision medicine enable prediction of how individual patients will tolerate specific agonists. Given that GI side-effects remain a meaningful barrier to adherence with GLP-1 agonists alone (and likely, in the future, dual agonists), we see great value in work aimed at improving patient experiences with tolerability in addition to efficacy. While we understand this is a relatively far ways away, the prospect is undeniably exciting.

Symposium: PIONEER Trial

Commentator

Cliff Bailey, PhD (Aston University, UK)

Dr. Cliff Bailey provided some fascinating commentary on Novo Nordisk’s PIONEER program for phase 3 oral semaglutide, weighing the pros and cons of an oral delivery route for GLP-1 agonists, plotting oral semaglutide’s impressive A1c reduction and weight loss vs. its injectable competitors, and positing a more “physiologic” mechanism behind its efficacy. In an impressively concise and illuminating figure (below), Dr. Bailey illustrating that oral semaglutide is the second-in-class GLP-1 agonist for both weight loss and A1c reduction when used as a monotherapy (at each drug’s highest dose), falling only behind its injectable counterpart, Ozempic. Notably, the A1c reduction is flat for oral semaglutide regardless of concomitant medications, a trend that also seems to hold more strongly for Novo Nordisk’s other GLP-1s, Ozempic and Victoza, than other GLP-1s.

What exactly contributes to oral semaglutide’s efficacy? According to Dr. Bailey, the oral delivery method possibly confers a more physiologic mechanism compared to injection. As it is digested in the stomach, oral semaglutide enters one’s system proximal to the liver, in a manner far more similar to human GLP-1. This confers a greater likelihood, as Dr. Bailey sees it, of triggering the GLP-1 receptors on the cells in the portal system responsible for directing blood from parts of the gastrointestinal tract to the liver, in turn activating the vagal afferents (nerves with sensory properties) which stimulate hypothalamic mechanisms. This (i) enhances the satiety-promoting effect of GLP-1s; (ii) potentiates the delay in gastric emptying; (iii) increases pancreatic glucose-induced insulin secretion; and (iv) decreases pancreatic glucagon secretion – whew! This sounds similar to the GLP-1 discussions back when the compounds were in phase 3, though accentuated - we appreciated that Dr. Bailey simplified this “interesting root effect” in a cartoon, pictured below.

Further, Dr. Bailey noted that this mechanism may also decrease the proportion of GLP-1 agonist reaching the heart vs. injectable GLP-1s, which could alter the cardiovascular effect of oral vs. injectable semaglutide. For reference, the latter conferred an impressive 26% reduction on three-point MACE (HR=0.74, 95% CI: 0.58-0.95, p=0.02 for superiority) in the SUSTAIN-6 CVOT. Needless to say, we cannot wait to see the results from PIONEER-6, oral semaglutide’s CVOT, which was estimated to complete just last week (September 27, 2018) on ClinicalTrials.gov – topline results are expected within 2018. Of note, although this CVOT was not powered for superiority, the results will still be very interesting to see.

• As for the downsides of oral semaglutide, Dr. Bailey pointed to the fasting requirements, nausea, and the possibility of having to change other medication schedules as particularly noteworthy. The first of these, he stated, has a huge impact on the bioavailability of oral semaglutide. The SNAC carrier molecule that enables oral dosing facilitates faster absorption of the semaglutide molecule through the stomach wall (gastric mucosa), but this still takes 30-60 minutes. Dr. Bailey practically noted that this means eating breakfast 30-60 minutes later (or an earlier alarm in the morning). We’ve heard many attendees bring this up during Q&A at recent meetings, and while study investigators have generally emphasized that it hasn’t been a big issue for study participants, in the “real world,” it’s hard to tell what patient preferences may be.

• Looking to the future, Dr. Bailey identified four things we still “need to know” about oral semaglutide and four things we would “like to know.” For the former, he mentioned (i) durability of oral semaglutide’s efficacy; (ii) long-term efficacy (especially vs. comparators); (iii) CVOT results; and (iv) effect on renal outcomes. For the latter, Dr. Bailey listed (i) the mechanistic benefits of oral uptake (potential for NAFLD?); (ii) whether good responders could be pre-identified; (iii) the feasibility of fixed-dose-combination tablets; and (iv) cost.

Frontiers of Drug Development: Overcoming the Barriers of Oral Peptide Delivery

Daniel Drucker, MD (University of Toronto, Canada)

The always great Dr. Daniel Drucker offered interesting scientific and historical context on the development of oral semaglutide and non-injectable peptides more broadly, in a well-attended symposium that reflected immense interest in this phase 3 candidate from Novo Nordisk. Describing the hostile environment drugs are sent into – a body full of enzymes, mucosa, and tight cell junctions designed to protect the body – Dr. Drucker laid bare the significant barriers to delivering delicate proteins and peptides through the GI tract. Within a short span of time, a drug has to get through the mucosal layers of the GI tract, make its way into a cell, cross the basement membrane, and reach circulation: Oral peptide delivery, he says, is a fantastic challenge. It must account for stability, degradation, immune response, and bioavailability. To enable oral administration of semaglutide, the oral semaglutide formulation uses a SNAC carrier. In Dr. Drucker’s description, SNAC is a “beautiful carrier through the stomach mucosa,” providing for rapid passage that allows semaglutide through but doesn’t keep the mucosa permeable for too long (which could let unwanted materials through). It takes ~57 to 85 minutes for full erosion of the semaglutide tablet in the stomach, he says, but accumulation of semaglutide in venous circulation occurs within minutes of administration, reflecting the speed of absorption. SNAC actually causes a local increase in pH to improve drug solubility and protect from proteolytic degradation, and it plays multiple roles in the dosing of oral semaglutide. SNAC not only serves to increase permeation of semaglutide into circulation – it also increases intracellular uptake of semaglutide, which Dr. Drucker characterized as “rapid and transient.” Ex vivo experiments on rat gastric mucosa demonstrate the rapid increase in cell permeability upon addition of SNAC (within, at most, ten minutes), but the increased permeability is also completely gone within the hour. In the end, Dr. Drucker emphasized that the pharmacokinetics of oral semaglutide are not significantly delayed from what is seen with injectable GLP-1s. All of this, he emphasized in closing, comes in a tablet a little over a centimeter long – “not very different in size versus many other tablets we’re prescribing.”

Oral Presentations: Novel Drug Therapies: Moving Beyond GLP1

MEDI0382, A Dual GLP-1 Glucagon Receptor Agonist, Promotes Rapid Glucose Control and Significant Weight Loss in Patients with Type 2 Diabetes

Victoria Parker, PhD (MedImmune, Cambridge, UK)

In a riveting oral presentation, MedImmune’s Dr. Victoria Parker provided tremendously positive phase 2a efficacy results for AZ/MedImmune’s GLP-1/glucagon dual receptor agonist MEDI0382, including an exciting first-look at emerging time-in-range data. The 49-day study (n=65) randomized participants into two cohorts, each pitting MEDI-382 vs. placebo. In cohort one, FreeStyle Libre Pro was worn across the entire trial and participants were titrated more quickly; in cohort two, there was no Libre use and participants were titrated more slowly – see below. All participants had type 2 diabetes and an A1c ≥6.5% and ≤8.5; the primary and secondary outcomes presented only considered cohort one.

On the first co-primary endpoint, MEDI0382 demonstrated a significant (p=0.02) weight loss of 3.4 kg (~7.4 lbs) over the 49-day trial, compared to 0.1 kg with placebo (below). Dr. Parker was quick to point out that this weight loss began immediately and did not plateau after 50 days. However, weight did jump back up immediately following removal of therapy, which she attributed to the study’s short duration and the fact that no diet or lifestyle interventions were recommended to patients after the study.

On the second co-primary endpoint of percent change from baseline to end-of-study glucose area-under-curve (AUC) in the first four hours after a mixed-meal tolerance test (MMTT), 300 μg (the highest dose) of MEDI0382 demonstrated a significant (p<0.001) 22% reduction. This was compared to a 6% increase in the placebo group. Notably, a pooled analysis of both cohorts on this endpoint at the lowest dose (50 μg), after the 7-day titration (see above), exhibited an equally significant 29% reduction compared to a 2% drop with placebo. To our understanding, this means that an ~equal percent reduction in hyperglycemia was found with 50 μg and 300 μg of MEDI0382 compared to placebo, suggesting this particular effect may not be dose-dependent. This postulation is supported by the steady difference in the mean daily glucose levels throughout the trial, seen in the CGM figure below.

• A1c dropped a significant 0.7% (vs. 0.1% with placebo, p<0.001 for comparison) in cohort one and 0.8% (vs. 0.4% with placebo, p=0.037) in cohort two. This was from already-low baselines of 7.3% for cohort one and 7.5% for cohort two, and we view this as very positive for such a short duration. One audience member noted that these low baseline A1c’s presumably translate to relatively low insulin resistance in these groups – an important variable to consider. Similarly significant reductions in fasting plasma glucose were also found (below).

• Those randomized to MEDI0382 in cohort 1 experienced an additional ~25% absolute time-in-range – defined quite stringently as 70-140 mg/dl – vs. placebo and over the course of the study (68% vs. 43% of a 24-hour day, p=0.002 for comparison). What’s more is that Dr. Parker assured the audience that there was no increased time in hypoglycemia – wow! We really think this is an incredible result (and we’re wondering about 70-180 mg/dl), particularly for patient quality-of-life. Dr. Parker also suggested that the glycemic benefits and reduced variability offered by MEDI0382 may begin within hours of the first dose, as demonstrated in the decreased mean glucose levels in the pop-out CGM tracing below.

• On tolerability – which has really emerged as the Achilles heel of dual-agonists, as we see it – MEDI0382 results were somewhat less inspiring, but not severely so. In cohort one, 22 members (65%) of the therapy group experienced a treatment-emergent adverse event (TEAE), compared to 6 (46%) in the placebo group; 19% and 12% of those on MEDI0382 in the first cohort experienced nausea and vomiting, respectively, compared to 0% with placebo for both. Safety/tolerability was similar in cohort 2, despite a more gradual titration; 15 (85%) of participants randomized to MEDI0382 experienced a TEAE compared to 3 (50%) in the placebo group; 35% and 20% of those on the dual-agonist experienced nausea and vomiting, respectively, again compared to 0% with placebo for both. One injection-site reaction and one viral/respiratory tract infection with MEDI0382 led to study discontinuation, but diarrhea was notably low (<5% of those on MEDI0382, according to Dr. Parker). Importantly, Dr. Parker noted that longer titration of MEDI0382 in the second cohort did not significantly improve the candidate’s tolerability profile. Indeed, we find it important to note that nausea and vomiting actually increased in the more gradual MEDI0382 titration group (cohort 2), though the small sample size in this study really limits strong conclusions on tolerability.

• With results from Lilly’s GLP-1/GIP dual agonist LY3298176 also presented today at EASD (above), we thought it might be interesting to compare the available data for each (while acknowledging that they are inherently different molecules). At the highest dose (15 mg), Lilly’s candidate demonstrated extraordinary 11.3 kg (~25 lbs) weight loss over 26 weeks – about four times longer than this 49-day study of MEDI0382. Assuming that MEDI0382 continued the weight-loss trajectory observed in this seven-week trial (a huge assumption, granted), the candidate could give very similar mean weight loss. For sure, this conjecture makes many assumptions, but Thursday’s presentations made us very excited about the potential of dual agonists, particularly for weight loss. It’s worth noting that side-by-side comparisons point to greater weight loss, faster with the Lilly molecule, at least in the early weeks of the phase 2b study above (below). Both molecules, we think, represent a huge leap forward in incretin therapeutics, diabetes treatment, and even obesity pharmacotherapy – the hype is shaping up to be real. Here’s an approximate side-by-side comparison of weight reduction in the first few weeks of each study:

• If you’re looking for more information on this exciting dual-agonist, check out this presentation on the candidate’s promising efficacy in reducing hepatic fat content in type 2 diabetes (indicating potential in NAFLD/NASH!), and this presentation demonstrating a very significant (p<0.0001) reduction in glucose AUC compared to placebo. See our GLP-1/glucagon competitive landscape for more context on this exciting emerging therapy class, which has implications for type 2 diabetes as well as obesity and NASH.

Corporate Symposium: Tomorrow’s Management of Diabetes, Today (Sponsored by AZ)

Keynote: Modern Management of Type 2 Diabetes – Glucose Lowering is Not Enough

John Buse, MD (University of North Carolina, Chapel Hill, NC)

The venerable Dr. John Buse was given free rein to conclude AZ’s daylong symposium, focusing on the diabetes therapies currently in development that he finds most exciting. According to Dr. Buse there were 560 products in active development for diabetes as of 2017. Of these, 152 were new, never-before investigated molecules and 95 were investigating a novel mechanism. These are the candidates he singled out as most exciting:

• Oral semaglutide: An unsurprising choice, given the overwhelmingly positive results released so far for this first-ever oral GLP-1 agonist. Seven of 10 phase 3 PIONEER trials have read out, including PIONEER 1 (vs. placebo), PIONEER 2 (vs. Lilly/BI’s SGLT-2 inhibitor Jardiance), PIONEER 3 (vs. Merck’s DPP-4 inhibitor Januvia), PIONEER 4 (vs. Novo Nordisk’s own Victoza), PIONEER 5 (in moderate renal impairment), PIONEER 7 (vs. Merck’s Januvia, with dose adjustment), and PIONEER 10 (vs. Lilly’s GLP-1 Trulicity in Japanese patients). In particular, Dr. Buse highlighted PIONEER 1 in his presentation, emphasizing that mean A1c fell below 6.5% in the 14 mg group without increased risk of hypoglycemia – quite remarkable, in his opinion. We’ve been following this candidate since 2013, and our excitement certainly matches Dr. Buse’s.

• Imeglimin: Interestingly, this phase 2/3 candidate blocks oxidative phosphorylation, which Dr. Buse noted is critical to human function: “I’m not sure how they are going to get away with this one.” We, too, have been curious precisely how this mechanism works safely. Right now, the candidate is entering three pivotal trials (n=1,100 total) for the treatment of type 2 diabetes in Japanese patients (under Sumitomo Dainippon Pharma); Imeglimin is phase-3 ready in the US and EU (under Roivant), but the company will conduct more differentiation studies, especially in CKD, before advancing to phase 3 in 2019. A promising ~1.0% reduction in A1c over 24 weeks was identified, at a high dose of 1,500 mg, in a phase 2 trial (p<0.0001 vs. placebo). Moreover, Dr. Buse pointed to a study (n=170) demonstrating an ~0.7% decrease in A1c with sitagliptin + Imeglimin compared to sitagliptin monotherapy. Down the line, we’d be curious about combinations of Imeglimin with SGLT-2 inhibitors or oral GLP-1 agonists, and our sense is that Roivant especially is most interested in pursuing this candidate for effective glucose-lowering in patients with impaired renal function.

• TTP399: In the phase 2b AGATA trial, vTv’s hepatoselective glucokinase activator (800 mg) demonstrated comparable A1c change to sitagliptin (100 mg) over the course of six months, while also conferring greater weight loss, in patients with type 2 diabetes (below). To our understanding, though, vTv has not pursued the candidate further in type 2 diabetes. As highlighted on the company’s 2Q18 call, patient dosing for phase 2 of the JDRF-sponsored phase 1/2 SimplicT1 Study (n=126) of TTP399 in type 1 diabetes recently began. The study will examine change in A1c at 12 weeks in patients with type 1, and expected completion is January 2019. Preliminary phase 1 results (n=5) in this extremely small study were presented at ADA 2018, demonstrating that TTP399 was well-tolerated (no severe hypoglycemia or DKA), glycemic control was improved, and insulin bolus levels were reduced. We’re not sure what the company’s plans are for type 2, and vTv also has an oral GLP-1 agonist in phase 2b.

• Cell-based therapies: Dr. Buse pointed to three initiatives in particular: (i) ViaCyte’s PEC-Encap – fully-encapsulated, indirectly-vascularized, stem-cell derived islets (two-year data from the STEP ONE trial demonstrated formation of viable mature insulin-expressing cells, with some persisting for two years, but issues arose with foreign body response); (ii) DRI BioHub – a bioengineered mini-organ (first patient treated in a phase 1/2 clinical trial has demonstrated stable glycemic control without exogenous insulin and without episodes of hypoglycemia) and; (iii) CLBS03 therapy, utilizing immunoprotective T_regs³ (T-Rex, a phase 2 trial, is underway to evaluate the safety and efficacy of CLBS03 as a treatment for type 1 diabetes). See more on cell therapy for type 1 in our type 1 cures competitive landscape.

• While the forefront of diabetes therapy is certainly exciting, Dr. Buse also stressed that implementation of what we already have in SGLT-2 inhibitors and GLP-1 agonists must come first, stating that current under use of these highly efficacious therapies is, “frankly an embarrassment.” Indeed, very few people are on these therapies compared to how many could benefit. A 2016 analysis found that SGLT-2s and GLP-1s accounted only for a collective 14% of all second-line type 2 diabetes prescriptions in the US, while sulfonylureas comprised a shocking (but declining, in recent years) 46%. Hopefully providing a much-needed tailwind for these therapies will be data demonstrating their efficacy in those without a history of CV disease, which Dr. Buse highlighted could come from DECLARE (AZ’s dapagliflozin, in a ~60% primary prevention cohort) for SGLT-2 inhibitors. And, we’re curious about the potential for REWIND (Lilly’s dulaglutide) to show cardioprotection with a GLP-1 agonist in a majority primary prevention population. Lastly, much more also needs to be done on access and patient affordability, in our view.

• To conclude, Dr. Buse extolled the digital ecosystem for using real-world patient data to improve diabetes management (CGM, Medtronic/IBM Watson’s collaboration, and consumer-driven technology were given as three examples) and delineated the implications that digital health has for clinical trials. As he sees it, digital health can (i) streamline patient recruitment to trials via registries by preselecting the most eligible candidates; (ii) reinforce patient engagement with trials; (iii) eliminate unnecessary clinic visits or bring the trial to the patient’s home; and (iv) support primary and secondary clinical trial endpoints (e.g., using medical devices to provide specific endpoints). Notably, we are already seeing these implications in the real-world. Science 37, a fascinating new company, enables “site-less” clinical trials – a trend we hope to see expand in diabetes as connected glucose monitoring and insulin delivery devices proliferate, enabling passive data collection and a whole new era of evidence generation at scale. Specific to diabetes, Science 37 is listed as a site on ClinicalTrials.gov for Lilly’s currently-recruiting smart pen trial investigating missed meal boluses. Science 37 is also enrolling a diabetes probiotic study in type 2, as well as a NASH study. 

Posters

MEDI0382, a GLP-1/Glucagon Receptor Dual Agonist, Reduces Weight and Improves Metabolism via Central and Peripheral Actions

D Baker, S Oldham, S Will, C Church, G Davies, L Burke, L Brown, M Sulikowski, JM Lapointe, A Lewis, H Jouihan, P Barkholt, R Veggerby Grønlund, C Rhodes, and J Trevaskis

A poster detailed results of a phase 1/2 dosing study for AZ’s GLP-1/glucagon dual agonist MEDI0382, unveiling a solid profile in terms of glycemic control, weight loss, and tolerability. Participants (n=61) were randomized to receive daily injections containing 100, 150, 200, or 300 μg MEDI0382 or placebo for 8 to 22 days. On the glucose front, all MEDI0382 doses produced marked ~40% reductions in both fasting and post-prandial glucose (p<0.0102). Additionally, all doses produced weight loss, and this reached statistical significance for the 300 μg dose (-3.4 kg [7.5 lbs] vs. -0.7 kg [1.5 lbs] with placebo) – a notable achievement after only 22 days on-treatment. Treatment-related adverse events occurred more often with MEDI0382 than with placebo (86% vs. 47%), but most events were mild or moderate. As is typical for a GLP-1 based agent, GI complications were the most commonly reported adverse events: Across the MED0382-treated groups, 27%-50% of participants experienced nausea (vs. 5% on placebo) and 8%-43% of participants experienced vomiting (vs. none on placebo). We’re looking forward to more data on dual agonists, particularly Lilly’s much-hyped, newly phase 3 GIP/GLP-1 this Thursday, and we’ll keep a particularly close eye on the balance of efficacy and tolerability. So far, it seems likely that these emerging dual agonists will carry more GI side effects overall, but it’s also important to balance this against expected greater efficacy on both glucose-lowering and weight loss. 

• In our view, these results bode well for the outcome of a larger phase 2b dose-ranging trial (n=834) of MEDI0382 (expected in 1H19), which according to AZ’s 2Q18 update will be a critical factor in the management’s decision whether to advance the candidate to phase 3. The candidate is being considered for multiple indications including type 2 diabetes and NASH, and no fewer than six phase 2 investigations are currently ongoing to further evaluate the candidate; we’re particularly intrigued by a study of MEDI0382 in combination with SGLT-2 inhibitor dapagliflozin (completion expected December 2018). Though the former is injectable and the latter oral, we’re hopeful about the combined weight loss and glycemic benefits these agents could provide. All this said, given AZ’s heavy investment in the candidate, we view it as a very reassuring to have these solid dose-ranging findings in the books.

Type 1 Cures

EASD/JDRF Symposium: Looking into the Future: Developing New Interventions for Type 1 Diabetes

Prevention Trials for Type 1 Diabetes: The TrialNet Experience and Looking into the Future

Carla Greenbaum, MD (Benaroya Research Institute, Seattle, WA)

The always-engaging Dr. Carla Greenbaum provided an update on the future of the TrialNet Program and urged clinicians to welcome the potential of immunotherapies in type 1 diabetes. We were especially intrigued by how Dr. Greenbaum framed the need for better language around treatment and prevention of type 1 diabetes. She offered up the example of how patients who have hypertension – who are therefore at a higher risk of stroke – are correctly treated for their hypertension in order to prevent strokes. These patients are not slapped with the label of “prestroke” and left alone. Dr. Greenbaum urged the audience to apply this same logic to type 1 disease progression: The early stages of type 1 (e.g. with 2+ autoantibodies but no glucose intolerance) should be viewed as opportunities for treatment in order to prevent progression into late-stage type 1 diabetes. We’d point out that this way of thinking should also be applied to the prediabetes/type 2 diabetes spectrum as well, as prediabetes is severely under-recognized as a disease state and promoting treatment in this population would be enormously beneficial (of course, Dr. Greenbaum’s work is squarely focused on type 1!). So, how can we best treat these early stages of progressive disease? First, the community needs to think beyond insulin treatments for type 1 therapy: Dr. Greenbaum chastised the status quo of type 1 treatment, commenting that “we are all too complacent with how we see [type 1] patients in 15-minute meetings and just tweak their insulin doses.”  Dr. Greenbaum touted the established benefits and potential of immunotherapy as a promising avenue for both type 1 treatment and type 1 prevention, and she advocated for endocrinologists to start adopting these therapies. As she explained, so many other autoimmune disorders are effectively treated with immunotherapy; why can’t the same apply for type 1 diabetes? To this end, she highlighted a number of immunotherapy studies currently underway, including a promising sequential study of rituximab followed by abatacept treatment that builds on previous studies showing their efficacy when administered separately.

• TrialNet is also now enrolling participants in a study investigating hydroxychloroquine (HCQ) in stage 1 type 1 diabetes, with the goal of delaying or preventing progression into stage 2/3. HCQ is a well-established agent, having been approved by FDA decades ago for the treatment of other autoimmune diseases, such as rheumatoid arthritis and lupus. Moreover, it’s also approved for use in children, streamlining its potential in type 1 therapy. The study, which is the first of its kind to investigate HCQ in diabetes, was motivated by previous data demonstrating HCQ’s ability to reduce innate and active immune activation in other autoimmune disorders; could a similar mechanism also work to slow the progression of type 1 diabetes? Per ClinicalTrials.gov, the trial is expected to complete in August 2024.

Hope and Hype: Where are We with Type 1 Diabetes?

Jay Skyler, MD (University of Miami, FL)

Building off a paper he published in Diabetologia earlier this year, University of Miami’s Dr. Jay Skyler tackled a fascinating topic: hope vs. hype in type 1 diabetes. The first half of his talk focused on “hype” with a tone of sincere frustration, followed by four “ideal therapeutic goals” in type 1 diabetes: (i) prevent immune destruction; (ii) preserve beta cell mass; (iii) replacement or regenerate beta cells; and (iv) automated insulin delivery. He concluded with promising research paths forward, highlighting the very cool DIPIT trial testing combination therapy in new onset diabetes – it has FDA and ethics committee go-ahead, but companies are making it difficult to get the anti-TNF and IL-2 drugs.

• Hope: Dr. Skyler emphasized that an “aggressive combination strategy” will probably be needed in type 1 diabetes, as there are there are many pathways that can be activated to target the beta cells. Therapies will ideally hit more than one pathway at a time, aiming to improve immunity (innate, adaptive, regulatory) and beta cell health.

• Dr. Skyler has proposed a combination strategy using anti-inflammatory agents, immunomodulation, agents that increase T-regulatory cells, diabetes-related antigen, and agents to preserve beta cell health. The Diabetes Islet Preservation Immune Treatment Trial (DIPIT) aims to use just such an approach, and it has notably received FDA IND approval and ethics committee approval to proceed – testing ATG, GCSF, IL-2, Etanercept, and Exenatide in new onset type 1 diabetes. Unfortunately, companies have not been willing to supply anti-TNF or the IL-2 drugs – airing concerns that other drugs will be used alongside their drugs (“My drug is going to get blamed if something goes wrong”). Dr. Skyler said it has been “very difficult” to navigate this, but the trial will hopefully get started. The ClinicalTrials.gov post currently has a start date of December 2018 and a primary completion date of January 2020.

• Dr. Skyler briefly touched on automated insulin delivery, noting “we are getting close” and “should see real progress coming in the next 24-36 months.” See our AID competitive landscape for a summary of the systems, which are mostly converging on ~2020 launches, with Tandem’s Control-IQ expected next summer.

• Hype: Dr. Skyler showed example after example of overstated headlines in type 1 diabetes research. These included Dr. Doug Melton’s work on betatrophin (retracted last year), the City of Hope’s 2017 goal to cure type 1 diabetes in six years, Andromeda’s DiaPep277, Dr. Denise Faustman’s phase 1 BCG vaccine study (see our July coverage), and many more. Dr. Skyler noted that over 400 interventions have cured type 1 diabetes in mice so far, and most type 1 cure headlines fail to appreciate this fact – mouse studies are just one link in a long research chain. He urged the field to think more carefully about how it articulates research findings, including for device – Dr. Skyler praised Medtronic’s cautious wording around the MiniMed 670G when it came out, but then lamented JDRF’s decision to use the term “artificial pancreas.” (This was impressively candid as JDRF CEO Derek Rapp was a session chair.)

• Dr. Skyler seemed most frustrated by the website PreventT1D.org, which urges people to take vitamin D, Omega 3 fatty acids, and a few other over-the-counter supplements to prevent type 1 diabetes – see the “cocktail” protocol here. Dr. Skyler noted that sites like these and social media “can create fake news” in the type 1 diabetes cure landscape. While some of these components are being tested by TrialNet, the claims made on a site like this are untested.

Additional Topics

Symposium: Management of Hyperglycaemia in Type 2 Diabetes: ADA-EASD Consensus Report 2018

ADA/EASD Consensus Report 2018

Melanie Davies, PhD (University of Leicester, UK); Apostolos Tsapas, MD (Aristotle University of Thessaloniki, Greece); Judith Fradkin, MD (NIDDK, Silver Springs, MD); Peter Rossing, MD (Steno Diabetes Center, Copenhagen, Denmark); Deborah Wexler, MD (MGH, Boston, MA); Geltrude Mingrone, MD (UNICATT, Milan, Italy); Walter Kernan, MD (Yale University, New Haven, CT); David D’Alessio, MD (Duke University, Durham, NC); Chantal Mathieu, MD (UZ Leuven, Belgium); John Buse, MD (University of North Carolina, Chapel Hill, NC)

In a packed 8 AM session, the ADA/EASD consensus committee presented its final consensus document on hyperglycemia management (Diabetes Care; Diabetologia), calling for “a paradigm shift” to more patient-centered care for type 2 diabetes and making groundbreaking recommendations for CVD, heart failure, and CKD. This document is an update to the associations’ 2015 position statement. We see the adoption of these guidelines as critical in multiple respects, and also missing in a couple of other ways – we hope, always, that HCPs in the US and EU will be supported as they move to implement more personalized and modern care for patients of all kinds – and that they have the right tools to do so (CGM, for example, isn’t mentioned at length even though it is covered by Medicare for those on MDI – we’d love to see Medicare cover CGM for those on SFUs).

The impressive report draws on 479 papers published over the past four years, offering recommendations on providing patient-centered diabetes care and choosing diabetes medications. Importantly, while Dr. Chantal Mathieu emphasized that this statement is not “guidelines” in and of itself, ADA Chief Science, Medical and Mission Officer Dr. Will Cefalu explained that ADA’s Standards of Care – now a “living document” – has incorporated this document and annotated Standards of Care with this new information. That these recommendations are explicitly emphasizing patient-centricity is a major positive. Patient preferences are fundamentally factored into treatment decisions; additionally, different adherence rates with various medications – and the impact that can have on outcomes – is a prominent factor to consider. Cost has also become, in our observation, a much bigger point of concern. We also perceive a strong focus of “closing the loop” per se; Dr. Cefalu described how patient-centered care is an ongoing cycle (below) in expressing his approval of the statement’s recommendations. The repetition is key: The report suggests patient and provider should work on a new plan at least twice a year, in order to better match evolving treatment needs. We love to see this and would encourage more thought on professional CGM and how that could be important in these meetings so that the blind aren’t leading the blind. Although we’d like to see a bigger technology orientation to help patients better guide themselves, we are grateful to EASD and ADA for working together for what must be thousands of collective hours to bring this together.

 

Additionally, we find this report to be a truly remarkable example of collaboration. We love that two major organizations put this together very collaboratively, encouraged and were receptive to review from the entire diabetes community, and received input over multiple months from a multitude of stakeholders. This included invitations to 50 reviewers and 870 comments from the public sent to the ADA alone! As detailed below, some truly meaningful revisions have taken place since presentation of the draft statement at ADA 2018 – it’s clear the committee took feedback very seriously. 

• This new guidance report provides detailed algorithms for choosing therapy, based on a patient’s primary area of concern. Specific roadmaps exist for choosing treatment based on ASCVD/CKD status, hypoglycemia minimization, weight loss concern, and cost issues. These are designed to reflect health history, weight, costs of care, and individual preferences. This primary emphasis on determining the right choice of diabetes medications is in turn supplemented by lifestyle management, self-management education, and external support. We are absolutely thrilled to see so much focus on so many respects of patient care and we are hopeful that “expert opinion” in the future will be used so that technology (especially CGM for those on SFUs and MDI) can enable better health.

• The statement explicitly recommends prescribing an SGLT-2 inhibitor or GLP-1 agonist for patients with atherosclerotic heart disease (ASCVD). If a patient is not meeting A1c goals, consideration should be given to adding an SGLT-2 or GLP-1 with proven CV benefit; if a patient is at target A1c, consideration should be given to switching to or adding one of these (as per the algorithm below). The algorithm recommends either an SGLT-2 or GLP-1 with proven CV benefit for those with predominant ASCVD, but prefers an SGLT-2 over a GLP-1 if heart failure or CKD predominates (SGLT-2s should only be used if eGFR is adequate). If A1c remains above target after this addition, or if the patient cannot tolerate the therapy, adding the other class should be considered; subsequently, any class with CV safety can be considered.

  • This section has seen notable revision since the draft statement was presented at ADA 2018. First, the equal recommendation of SGLT-2s vs. GLP-1s on predominating ASCVD was revised from a recommendation of GLP-1s over SGLT-2s – that’s a big deal. Additionally, chronic kidney disease (CKD) was added as a consideration in this algorithm, with a preference for SGLT-2 use in the case that CKD (or heart failure) predominates. At ADA 2018, Dr. Silvio Inzucchi commented that the evidence for renal benefit with SGLT-2 inhibitors was as strong as the evidence for heart failure, leading him to question why the committee had – at that point – chosen to make a recommendation on heart failure but not CKD. We imagine this is the logic that led to this specific revision, and we’re thrilled to see the renal benefits SGLT-2s have now demonstrated making their way into recommendations. We will look forward to hearing more details on the impact of GLP on kidney disease.

  • The committee stuck with its drafted “ranking” of SGLT-2 inhibitors and GLP-1 agonists. Within the GLP-1 agonist class, Victoza (liraglutide) is preferred over Ozempic (semaglutide), which is preferred over Bydureon (exenatide); this ranking makes sense given semaglutide’s (Ozempic’s) strong efficacy, and weaker CVOT data from SUSTAIN 6 (for Ozempic/semaglutide), which isn’t as strong as from LEADER (for Victoza/liraglutide). For SGLT-2s, Jardiance (empagliflozin) is preferred over Invokana (canagliflozin), presumably for reasons of safety (i.e. amputation risk for Invokana/canagliflozin). We wonder how, once the DECLARE CVOT fully reads out, Farxiga (dapagliflozin) will factor into this equation.

• Dr. John Buse said outright that there were not nearly enough patients taking cardioprotective diabetes therapies, given the strong evidence some of these have for conferring cardiovascular risk reduction. It is terrific for HCPs in particular to see cardiovascular and renal risk factors discussed in depth – we hope this also translates into action.

 

• The report also recommends GLP-1 agonist as the first injectable medication for most adults with type 2, while SGLT-2 inhibitors are suggested for those with chronic kidney disease. Both of these represent groundbreaking recommendations in diabetes, and we continue to see this as a big step forward given (i) increasingly strong evidence of cardio- and renal-protection; and (ii) better long-term safety and efficacy. As noted, these medicines do not prompt hypoglycemia as long as they are not taken with insulin. This prominent position for GLP-1s and SGLT-2s is very important – and we hope it will drive changes in HCP behavior so that patients who need them are encouraged by informed clinicians.

  • The authors discuss how to initiate GLP-1 – “you’ll have to read everything,” said Dr. Chantal Mathieu, who emphasized that titration to the maximum dose is very important. Notably, speaking on basal insulin initiation, she also said, “please do not just initiate – also titrate.” She also stressed the importance of intensive glucose monitoring as an absolute requirement with mealtime insulin. Adding some clinical wisdom, she also emphasized that for those for whom GLP-1 monotherapy isn’t enough, it’s better for most patients to go onto basal insulin and then intensify with a GLP-1 agonist, rather than the traditional mealtime insulin.

  • We do wonder about what saying that GLP-1s should be the “first injectable” may imply to many people with type 2 that GLP-1 is exclusively a “later stage experience.” Or even that injectables “must” or “should be” late stage! On the contrary, many experts point out that GLP-1 has shown to be of great use early in disease progression, particularly for preservation of beta cell function. We very strongly agree with the recommendation that GLP-1 is an option preferable to basal insulin for most patients, and we also think GLP-1s could be preferable to many other therapies for many patients – and this will only become truer with the advent of oral GLP-1 and potentially implantable GLP-1 when that is ready. Notably, for some patients, GLP-1/basal combinations may be a better therapy than a straight GLP-1 – this was acknowledged by the leaders.

• We’d like to see CGM more strongly encouraged when patients have a meaningful risk of hypoglycemia. While we understand the fundamental need to be evidence-based, CGM is pretty heavily downplayed in the new guidelines – while of course it is true that there is not too much evidence (who is going to run a CGM trial in type 2s taking SFUs?), expert opinion is important. In our view, given that CGM is still very underutilized in people with both type 1 and type 2 diabetes, it would be great to see guidelines promote its use where its value can be reasonably assumed. There are so many patients on basal-bolus insulin therapy or taking insulins that carry hypoglycemia risk (not to mention SUs) who are at major risk of severe hypoglycemia – to say nothing of the tens of millions of patients who experience far too much hyperglycemia each day.

• From Professor Miles Fisher (Glasgow Royal Infirmary, UK): "I would comment that this is a very detailed document and that it requires detailed reading alongside looking at the presentation that was given. The suggestion of GLP-1 RAs first after tablets for the majority fits with current routine practice in the UK and is welcome. The figures are complex and will be difficult for primary care doctors to use so some form of simplification would help. The insulin figure in particular is hard to follow. There was an error when it was presented in that Peter Rossing said the HFH (heart failure hospitilization) was not adjudicated. He means that CHF (congestive heart failure) at baseline was not adjudicated. In all of the CVOTs, HFH was blindly adjudicated by a CV endpoints committee and the criteria applied were the same as those use in CV studies of heart failure treatments. As a broader comment I think that there should have been a cardiologist on the committee!"

Symposium: UKPDS

Putting the UKPDS into Perspective

David Mathews, DPhil (University of Oxford, UK)

In a standing-room only lecture, Oxford’s Prof. David Matthews methodically dispatched, in a quite impassioned manner, four misconceptions surrounding UKPDS that he says he and co-investigator Prof. Rury Holman hear all the time. The misconceptions were all fairly trivial and fallacious at the end of the day, which perhaps explains why Prof. Matthews was tired of hearing about them. We imagine that after this lecture, which was summed up with a wrap-up slide that left nothing ambiguous (each statement succeeded by “Wrong!”), Prof. Matthews may hear these challenges a bit less. Indeed, the 84-publication-strong UKPDS has driven clinical care guidelines for type 2 diabetes across the globe, and improved long-term outcomes substantially.

• Misconception #1: “We cannot trust the UKPDS results about intensive glucose control – another trial showed that lowering A1c to normal led to an increase in death rates.” The principle issue with this statement is that UKPDS was performed in a population with low diabetes duration (quite recently diagnosed) and low CV disease. Subsequent trials to wish the mis-informed may be referring, such as VADT, ACCORD, ADVANCE, and PROactive, enrolled cohorts with advanced duration and established CV disease, an entirely different risk population. Further, there were fundamental differences in therapy: UKPDS used diet and SUs or insulin (with insulin rescue as necessary), while ACCORD used a large array of glucose-lowering therapies and insulin with a treat-to-target scheme. Most type 2 trials since UKPDS do not address issues of early glycemic control.

• Misconception #2: “The UKPDS used old-fashioned glucose-lowering techniques which are irrelevant in today’s new drug environment.” The UKPDS was a “trial of policies – glycemic control in newly-diagnosed type 2 diabetes – not of therapeutics.” We’re somewhat shocked to hear this misconception since although it’s true that there are now glycemic-dependent compounds available, such a small percent of type 2 patients globally access them.

• Misconception #3: “The metformin arm of the UKPDS was underpowered using only 342 subjects – we don’t know that the result was true.” He responded simply that there was sufficient power because trial sizes are reported on the number of people randomized, not those in an individual arm. There were 342 people in the metformin arm, but 411 in the conventional therapy arm, making total n=753. In the DCCT (n=1,441), the split in primary prevention was n=378 vs. n=348.

• Misconception #4: “Many recent trials of new glucose lowering agents showed no difference in cardiovascular outcome – lowering glucose therefore cannot be very important.” FDA-mandated CVOTs are predicated on striving for glycemic equipoise to show that the agent is not unsafe compared to others that lower blood glucose to the same extent, isolating the impact of the molecule. “But,” Dr. Matthews explained, coining the glycemic equipoise paradox, “if we do not allow the drug to have the differential effect for which it was designed, then we don’t get the information we need. This would be like running a statin trial and pre-specifying that the control arm and the active arm needed to have the same cholesterol levels.”

Symposium: Diabetes and the Mind: Psychological Aspects

Glycaemic Variability: Does it Really Matter?

Thomas Danne, MD (Paediatrics, Children’s Hospital Auf der Bult, Hannover, Germany)

Prof. Thomas Danne lamented that “we doctors have done very bad things to patients for some time because of the DCCT,” and made a compelling case for widespread adoption CGM, time-in-range, and glycemic variability. He reviewed the progress to date in moving beyond A1c and achieving consensus on CGM outcomes, pointed to the (limited) evidence base currently available and asked for further studies, introduced a blood pressure-like way of reporting glucose in a post-A1c-centric world, and launched into emotional spiels regarding the emotional impact of glycemic variability. The next steps, which he said are happening in the US, but unfortunately not the EU, are to get regulators and health policymakers on board the beyond A1c train. FDA (and EMA) continues to engage openly with professional societies and the patient advocacy community, but there’s still clearly a long way to go before glycemic variability, glucose excursions, time-in-range, hypoglycemia, and PROs are widely adopted. But Prof. Danne ended on an encouraging note: “Clearly, A1c is on the way out.”

• “Do we have evidence this is related to long-term complications? Well, no. I think we need to do studies. We need to do longitudinal studies.” Prof. Danne touched on a number of existing studies, showing that time-in-range correlates with retinopathy, that glucose standard deviation correlates with alterations in the white matter structure in type 1 children, that MAGE correlates with variability in neointimal thickness and frequency of uncovered stent struts (CV markers), and that higher day-to-day fasting glycemic variability correlated with higher risk of severe hypoglycemia (and all-cause mortality) in DEVOTE. We’d add Dr. Roy Beck’s recent analysis, presented at ADA, showing that SMBG-derived “time-in-range” correlated with microvascular outcomes in the DCCT, to that list. Still, those studies are cross-sectional, observational, and/or don’t use CGM – they lack the rigor required of regulatory and payer bodies. The clear low-hanging fruit in validating CGM metrics, in our view, is to do so with the acute effects of hypoglycemia. It seems feasible to show in a prospective observational study that time spent <70 mg/dl is tied to greater occurrence of severe hypoglycemia, an outcome in which both FDA and payers have interest. The same study may also find correlations between glycemic variability and hypoglycemia, as a Bergenstal/Novo Nordisk poster did at EASD. Demonstrating links to long-term complications is an inherently more complex problem, and the field has all but agreed that a second DCCT, this time with CGM, is not necessary, too time consuming, perhaps unethical, and too expensive. Rather, we hear investigators talk of retrospective analyses (a la Dr. Beck’s) with existing data sets, as well as prospective observational studies with the use of biomarkers and risk prediction scores.

  • Prof. Danne raised the possibility – which he believes – that postprandial hyperglycemia has a unique role in causing complications. There is no definitive proof that this is the case, but it is similar to data that Dr. Rory McCrimmon shared at last year’s EASD, which suggested that hypoglycemia may be most deleterious in the context of hyperglycemia. This interplay will be interesting to study from the basic science perspective, as well as clinical, which could also leverage data from observational CGM studies/registries.

  • Prof. Danne briefly touched on the economic effects of postprandial hyperglycemia. In a 2016 web survey of 906 adults with type 1 or type 2, 561 reported experiencing postprandial hyperglycemia in the past week (we’re surprised this wasn’t even higher). Compared to those who didn’t have hyperglycemia in the past week, those who did performed nearly twice as many fingersticks, made twice as many calls/emails to HCP per year, and had 25% more HCP visits per year (5.5 vs. 4.4). One out of four reported missing work time. An EASD 2017 poster sharing 12-month healthcare outcomes from national reimbursement of CGM in 515 pump users, with an estimated nationwide cost reduction of €345,509 (driven by reduced hospitalization/absenteeism). These sorts of analysis will go a long way, particularly in payer conversations.

• We were intrigued by Prof. Danne’s suggestion to present CGM metrics in a manner similar to blood pressure, because “no one says their blood pressure is 120, they say 120/80.” Similarly, a time between 70-180 mg/dl of 52% doesn’t quite tell the whole story…what’s the hypoglycemia? So Prof. Danne advocated for routinely reporting time-in-range(s) as “52/15,” where the first number is the percent of time between 70-180 mg/dl and the second number is the percent time <70 mg/dl. This system of nomenclature makes good sense to us!

• Prof. Danne interspersed comments about the detrimental psychological effects of highly variable blood glucose:

  • “I think we have to shift our whole way of approaching diabetes, away from A1c and toward glycemic variability. Because we know that glycemic variability is having a big impact on behavioral and emotional feelings, that these glucose swings, glucose all over the place, makes people nervous, gives them a headache, they don’t feel well, it might lead to depression and anxiety, it might lead to a reduction in cognitive tests, it might have an economic effect because often people don’t go to work! All this will increase stress related to the disease.”

  • “Patients with extremely variable spaghetti plots say they are extremely stressed.”

  • “People experience frustration and learned helplessness. Why? Because there are no realistic goals (that’s the diabetes care team’s fault), too high expectations (from their parents), low tolerance to frustration, and insufficient knowledge. That can really lead to negative self-esteem, no social integration, and diabetes burnout.”

Symposium: Diabetes Prevention on Demand: What Strategies

Strategies to Implement for Diabetes Prevention

Rafael Gabriel, MD (University of Madrid, Spain)

Universidad Autónoma de Madrid’s Professor Rafael Gabriel unapologetically challenged calls-to-action for diabetes prevention from non-governmental organizations (WHO, IDF, ADA, EASD, GACD): “So what?” As he sees it, this ostensible support for diabetes prevention is ineffective and a call-to-action is not enough – more actual action is required. In order to impact public health pandemics as rampant as diabetes and obesity, many actors and stakeholders need to be recruited and involved, from industry to civil organizations to politicians: “To implement public health initiatives we need politics.” Sadly, pleas for political progress are met with the reality that politics today are more often than not characterized by partisan gridlock, even on issues central to nations’ health. If broad and comprehensive action doesn’t take place soon (incorporating politicians especially), Prof. Gabriel believes that we will still be talking in circles around the science and implementation of diabetes prevention in 20 years, with no progress to show for it. From his perspective, the major headwind to progress is the transfer of evidence to population-level implementation, which requires seamless communication and diligent work between investigators and politicians/other stakeholders in public health. To overcome barriers, Prof. Gabriel firmly stated that “sometimes we need to make decisions and implement policy based on imperfect evidence,” which will always be the case given the complex nature of diabetes etiology and its unfathomable scale. Accordingly, he believes that no further primary prevention trials are needed to support the evidence supporting use of DPPs. Pointing to seven major diabetes prevention studies (the Da Qing IGT Study, Finish Diabetes Prevention Study, US Diabetes Prevention Program study, Indian Diabetes Prevention Programmes 1 and 2, Zensharen Study, and the Prevention of Type 2 Diabetes by Lifestyle Intervention study), Prof. Gabriel asserted that evidence now unequivocally supports that effective diabetes prevention programs can give a 50% decrease in the risk of type 2 diabetes – a degree of efficacy unmatched by any diabetes drug currently on the market. Are diabetes prevention programs scalable enough to reach all 84 million people in the US, including the ~90% who don’t know they have it, without quickly reaching a point of saturation? Surely, there’s also a role for complementary population-level interventions such as taxes on sugary food.

• Following Prof. Gabriel’s passionate speech, session chair Dr. Paolo Pozzilli suggested a “diabetes and politics” session at future diabetes meetings – an intriguing prospect he suggested could be aimed at fundraising and increasing dialogue. We would be front row at this session! Other topics could address reimbursement and drug (insulin) pricing – anything to better inform and engage the scientists and providers on the front lines in policy-making!

• In response to the audience question, “Why not just ban sugar-based drinks?” Prof. Gabriel responded, “We certainly should learn from other fields. For instance, tobacco and smoking cessation and control. Of course, we need some sort of legal approach, perhaps with taxation or prohibition of tobacco, sugar beverages, fast food in schools and public spaces. These legal measures can mitigate the level of the pandemic, but we also need action at the community level.”

• According to Prof. Gabriel, lifestyle interventions must target, at a minimum, eight specific things: (i) reduction in weight of >5%, (ii) total fat intake less than 30% of energy consumed, (iii) saturated fat intake less than 10% of energy consumed, (iv) reduction of daily salt intake (<3 g/day; for CV health too), (v) fruit or vegetable intake of at least 500 g/day, (iv) moderate-intensity exercise of at least 30 min/day, (vii) smoking cessation, and (viii) mobile health motivational interventions through text messaging. Prof. Gabriel didn’t have a citation on the slide nor mention how previous study interventions have stacked up, so we’re not sure of the genesis of these recommendations. We did note the brief allusion to digital health’s importance in diabetes prevention. As CDC’s Dr. Ann Albright said in July, “We must utilize technology…we’ll never get to anything of scale unless we use digital.”

• Prof. Gabriel drew a distinction between two strategies to prevention implementation: population-level (e.g., soda tax) and high-risk (e.g., DPP). Cambridge’s Prof. Nick Wareham gave an incredible talk on the two approaches at WCPD in July. Population-level approaches aim to shift the entire “bell curve” of diabetes to the left, often leading to smaller risk reductions in larger numbers of patients, and in a less resource-intensive manner. In contrast, high-risk interventions target high-risk patients (as the name implies), begetting greater risk reduction put in smaller numbers of people – this approach is often more resource-intensive. Prof. Gabriel underscored the need to strike a balance between the two, arguing that focusing on too high-risk a population may prove futile for prevention, while focusing on too many people at the low-risk end of the spectrum could prove cost-ineffective. Interestingly, both approaches build on what Prof. Gabriel characterized as the still-controversial concept of a “spectrum of health,” where everyone lies on the continuum between healthy and diabetes, but there is no categorical distinction between the two. We think this concept is a whole lot less controversial than Prof. Gabriel seems to, especially in light of the growing body of evidence that prediabetes is a major CV risk factor in and of itself.

The Changing Face of Type 2 Diabetes Epidemiology

Jaakko Tuomilehto, MD (Dasman Diabetes Institute, Kuwait)

The esteemed Dr. Jaakko Tuomilehto offered sobering commentary on two troubling new trends in the diabetes epidemiology landscape: The rise of diabetes-associated cancer and Alzheimer’s disease. On cancer, Dr. Tuomilehto pointed to a hot-off-the-presses European Journal of Epidemiology study reporting a significantly increased risk of several different cancers among people with type 2 diabetes in his native Finland. These included lip cancer (IR=1.40, 95% CI: 1.28-1.53), liver cancer (IR=2.44, 95% CI: 2.35-2.53), pancreatic cancer (IR=1.75, 95% CI: 1.70-1.79), stomach cancer (IR=1.22, 95% CI: 1.18-1.26), colon cancer (IR=1.22, 95% CI: 1.19-1.25), gallbladder cancer (IR=1.29, 95% CI: 1.21-1.36), skin cancer (IR=1.18, 95% CI: 1.15-1.22), kidney cancer (IR=1.42, 95% CI: 1.37-1.47), bladder cancer (IR=1.17 95% CI: 1.13-1.21), and thyroid cancer (IR=1.22, 95% CI: 1.12-1.31). We’re curious whether the increased of risk cancer with diabetes is an artefact of the cancer risk associated with overweight and obesity (which are set to surpass smoking as the leading cause of cancer mortality), or whether dysglycemia itself plays an independent role in cancer risk – possibly, the underlying pathophysiology common to diabetes and obesity could be giving rise to increased cancer risk. This is an open question, but Dr. Tuomilehto forecasted that cancer will soon have a similar standing as CV disease as a full-fledged and incredibly urgent complication of diabetes. On Alzheimer’s disease, Dr. Tuomilehto also turned to Finnish registry data: In this population, 10 year incidence of Alzheimer’s disease and vascular dementia are 2.45-fold and 2.15-fold higher for people with diabetes, respectively. He noted that this alarming trend is likely to intensify in the coming years as people continue to live longer and longer with diabetes. However, on a hopeful note, Dr. Tuomilehto pointed out that cognitive function in older adults is responsive to lifestyle modification in the same way that diabetes and obesity are. Looking beyond epidemiology, we can’t help but notice that cancer and Alzheimer’s disease are two of today’s most well-funded and high-profile disease states. We can only imagine how much we could move the needle on the urgency of the diabetes epidemic if the oncology and gerontology communities emphasized diabetes prevention and management as a mitigating measure for cancer and Alzheimer’s disease risk.

Symposium: Diabetologia Symposium: Diabetes and Microbiota: Close to a Missing Link, or Ordure Up to the Ankles?

The Role of the Gut Microbiota in Human Type 2 Diabetes: Insights from Multi-Omic studies

Jordana Bell, PhD (Kings College London, UK)

Dr. Jordana Bell shared her vision of leveraging the microbiome for the prevention, risk stratification, and even treatment of type 2 diabetes; moreover, she had the data to back it up. Her team analyzed gut microbiota from 2,401 samples (including 2,280 controls) to discern differences between those with a type 2 diabetes diagnosis prior to sampling (“retrospective,” n=91) and those with a type 2 diabetes following sampling (“incident,” n=30). The unpublished data not only revealed a distinct diabetes cluster from the controls, but also a separation between patients with retrospective and incident diabetes. In fact, nine microbial genera were shown to be associated with incident type 2 diabetes. Dr. Bell characterized the results as “very consistent.” Dr. Bell hopes to eventually use these findings and others to apply an interventional approach for type 2 diabetes, likely via diet modulation. The basic premise, she explained, would involve an initial blood test of patients’ fasting glucose. Those who were determined at risk of diabetes would undergo gut microbiota profiling, after which targeted interventions might be applied to halt further deterioration and rising insulin resistance. To this end, Dr. Bell listed several potential treatments: (i) dietary modulation, which she deemed “the most realistic and tangible,” in the form fat-restricted or carbohydrate-restricted low-calorie diet; (ii) supplementation with prebiotics and probiotics to enhance bacteria strains known to alter gut microbiota beneficially; and (iii) fecal microbiome transplantation – she noted at least one trial showing this intervention to improve insulin sensitivity in peripheral muscle in men. These potential interventions are certainly intriguing alternatives, and industry has already begun to capitalize on growing public interest in the microbiome. In March, the Wall Street Journal featured DayTwo and Viome, two startups sequencing gut flora to provide personalized nutrition recommendations – while these offerings haven’t quite moved into the arena of therapeutic advice, we assume it will only be a matter of time as the science improves.

• Despite the exciting developments achieved thus far, much work remains. Dr. Bell called for a double-blind, randomized controlled trial investigating butyrate/short-chain fatty acids supplementation (known to play a role in diabetes) or fecal microbiome transplantation. Already, trials are helping to clarify the emerging field: one randomized controlled trial investigating a diet enriched for fiber intake identified that not all short-chain fatty acids matter for diabetes; rather, only a “very small subgroup” altered outcomes associated with alleviation of type 2 diabetes symptoms. Dr. Bell noted that these results are “very exciting.”

• While diabetes has been known to be associated with gut microbiota, Dr. Bell was surprised by results showing type 2 diabetes to have one of the highest associations with gut microbial markers out of 38 diseases investigated. In fact, type 2 diabetes was surpassed only by allergies for ingested foods and inflammatory bowel disease. Dr. Bell described the results as “quite promising,” as they suggest substantial therapeutic potential in modulating gut microbiota for type 2 diabetes.

Symposium: Deconvoluting Big Data: Seeing the Trees in the Forest

From Big Data to Clinical Application

Anders Rosengren, MD (University of Gothenburg, Sweden)

Dr. Anders Rosengren presented valuable insights into the potential to leverage new computational techniques and increasingly large datasets for drug discovery. In his research, Dr. Rosengren takes a big data-focused approach to identify genes involved in diabetes, and then -importantly - characterizes the identified targets through molecular and physiological studies, and eventually in human clinical trials. Dr. Rosengren stressed that big data alone is not enough – rather, “big data in conjunction with type 2 diabetes physiology can help us to find new mechanisms.” He has used this approach to identify multiple new candidates for type 2 therapies, which he has since validated in animal and human trials. In one fascinating example, Dr. Rosengren’s team used a “network model” to identify groups of interconnected genes associated with hyperglycemia in a mouse model. Given that diabetes can’t be traced to any single factor and results from interactions within many different physiological systems, this network approach is particularly valuable, as it helps to identify these connections. The investigators then cross-referenced this network with human type 2 gene variants to generate a list of the 50 genes most likely to be associated with type 2. By using a database of thousands of drugs and their effects on gene expression, the team was ultimately able to identify specific compounds that would have the greatest overall effect on the 50 type 2-associated genes. Surprisingly, the leading candidate that resulted from this analysis was sulforophane – a natural compound found in broccoli and other vegetables. The team didn’t stop there and went on to demonstrate that this compound decreased glucose production in cell lines, improved glucose tolerance in mice, and improved fasting glucose (-0.22 mM, p=0.023) and HbA1c (-3.7 mmol/L, p=0.034) in patients with obesity and dysregulated type 2 compared to a placebo group. Notably, the researchers think this drug could serve as a first-line alternative to metformin. Dr. Rosengren’s research approach serves as a valuable model for the field: both of the awesome power of big data to identify new avenues for treatment, as well as the importance of rigorously following-up these new connections in patient populations.  

EASD/AASD Symposium: Tackling Diabetes at a Whole System Level

Diabeter: Implementing Value-Based Healthcare in Paediatric Diabetes in the Netherlands

Henk Veeze MD, PhD (Founder and Medical Director, Diabeter, Netherlands)

Diabeter founder Dr. Henk Veeze described ongoing developments at the innovative Medtronic-owned chain of Dutch clinics with an advanced integrated care model that delivers value-based medicine. He shared that Diabeter, which started with just 500 patients in 2006, has now nearly 2,500 patients and is expected to break even financially in the next two to three years. We were very interested to hear details on Diabeter’s decision to pursue long-term (10 years) value-based healthcare contracts with insurance companies. Dr. Veeze noted that Diabeter has already signed such a contract with “the biggest insurance company” in the Netherlands. The rationale, Dr. Veeze explained, is for the clinic to invest now and save later, reducing future costs by receiving a percentage of the cost reduction annually. The additional funds can then be invested back into the clinic to continue improving patient care and outcomes. We think paying upfront for healthcare costs makes a lot of sense, but as one of our former summer associates pointed out during the Q&A, we wonder if the incentives for such a long-term contract exist in the bulk of the US, given that insurance churn is fairly high. Dr. Veeze responded that hopefully insurance companies would understand that regardless of whether patients move between companies, this kind of system would benefit the collective (a “pay-it-forward” model, where everyone participates). This outlook may be a bit optimistic for the US (despite Obamacare’s preventive services mandate), but we think there is still plenty of potential for value-based care contracts – perhaps just on a shorter time scale.

• Dr. Veeze intriguingly shared that in 2019, Diabeter will expand to “other countries” – we wonder if this might be a reference to type 2 Diabeter clinics in the US. On Medtronic’s 1Q18 call, a slide alluded to bringing the Diabeter model to US type 2 diabetes clinics. While the call did not share timing or further details, the potential for the Diabeter model, which so far has only been in type 1s, to come to the US in type 2s is very exciting. There is high potential and high need in the type 2 population, given the fragmented care many patients receive. Dr. Veeze stressed that he is “very open to teams who want to join and learn.”

• Diabeter’s bundle payment system calculates outcomes based on changes in quality of life and glycemic control – we love that both critical metrics are included. While Diabeter is 100% responsible for the bundle, costs are kept down by keeping suppliers mainly internal. 74% of bundle costs are incurred by Diabeter and Diabstore (the internal supplier) to provide medical devices and hospital care. External costs for hospital care, medical devices, and medication comprise just 7%, 8%, and 11% of the total bundle, respectively.

EASD/EASL Symposium: Diabetes and NAFLD

NAFLD: How to Diagnose and Monitor Patients

Laurent Castera, MD, PhD (University of Paris-VII, Hospital Beaujon, Clichy, France)

Dr. Laurent Castera reviewed new non-invasive tests to help avoid biopsies and outlined a diagnostic strategy for detecting NAFLD in patients with diabetes. New strategies to detect NAFLD are sorely needed, given its high prevalence and the current barriers to widespread screening. Dr. Castera explained that up to 70% of patients with type 2 diabetes are estimated to have NAFLD, making it a no-brainer that every patient should be screened. The current gold standard diagnostic tool is a liver biopsy, which Dr. Castera argued is not practical for routine and widespread screening. To illustrate this, Dr. Castera estimated that ~25% of France’s 67 million residents have some form of NAFLD, yet there are only 400 hepatologists in the country, and about 40 liver-specializing pathologists – there simply aren’t enough physicians qualified to perform and then assess liver biopsies. Furthermore, patients understandably don’t want the expensive and painful procedure and are wary of rare complications including uncontrolled liver bleeding and death. Fortunately, new non-invasive tools have become available that can help doctors detect signs of fibrosis so that they can recommend biopsies in a more targeted way. These include serum biomarker tests and FibroScan, which measures liver stiffness. The FIB-4 and NFS scores are the best-validated biomarkers for detecting advanced fibrosis. These biomarkers can be easily scored by a general practitioner, as they are based on commonly used blood tests and patient metrics, including the alanine aminotransferase (ALT) and aspartate aminotransferase (AST) tests, albumin concentration, platelet count, patient age, and BMI. High scores indicate referral to a hepatologist who can perform further tests and possibly recommend a biopsy. FibroScan uses ultrasound imaging to measure liver stiffness and can correctly identify cirrhosis at least 90% of the time. However, the test becomes less accurate as body mass increases, limiting its value in some of the most at-risk populations. Dr. Castera stressed that these tools aren’t accurate enough to positively identify NAFLD on their own, but can rule out severe fibrosis and indicate referral to a specialist; moreover, they are cheap and accessible, making widespread screening possible.

• Dr. Castera’s strategy for NAFLD screening in diabetes patients relies on increased cooperation between PCPs, endocrinologists, and hepatologists. For patients presenting with symptoms of liver disease, endocrinologists or PCPs can rule out non-NAFLD causes (such as viral hepatitis), and then check serum biomarker scores like FIB-4 and NFS. If these indicate high risk, patients can be referred to a hepatologist who can assess liver stiffness before deciding whether or not to biopsy.

• In the Q&A, it was pointed out that routine NAFLD screening is currently recommended in Europe, but not in the US. Dr. Castera attributed this to the much earlier approval of non-invasive tests in Europe and Asia than in the US, and asserted optimism that, as American physicians become more familiar with FibroScan and serum biomarker tests, more widespread screening might become indicated in the US. 

IDF Europe Symposium: People with Diabetes at the Centre of Care

People with Diabetes at the Centre of Care

Sehnaz Karadeniz, MD (Istanbul Science University, Turkey)

The IDF Europe symposium emphasized that people with diabetes must be at the center of care. Mr. Uros Bogdanovic, a 22-year-old law student with type 1 diabetes and member of the Serbian Diabetes Association, argued this point, explaining how patients have unparalleled knowledge regarding the ways in which their condition manifests in everyday life. Moreover, patient voices hold a “certain weight,” and young patients especially are motivated to instigate change. Still, Mr. Bogdanovic acknowledged that there are barriers to positioning patients as equal to other stakeholders, citing a lack of resources and preparedness. Additionally, Mr. Bogdanovic explained, patients can also be “selfish and want improvements just for themselves.” To this end, he described how the Serbian Diabetes Association successfully petitioned the government to increase the number of test strips provided to patients over the age of 26. In the past, those <26-years-old received 150 test strips/month, while those >26-years-old received just 50 test strips/month – that’s not even two strips/day! The ultimate compromise provided those <18 years-old with 150 test strips/month and those >18 years-old with 100 test strips/month, doubling the number of test strips allotted to adults. However, Mr. Bogdanovic noted, patients between the ages of 18- and 26-years-old were unhappy with the decision, as they no longer received 150 test strips. Given the complexity and diversity of diabetes, unifying patients can be tricky, especially when unintentionally harmful language enters the equation – more on that below. Read on for further insights from each of the three compelling panels on patient advocacy, language, and health systems.

• In a panel focused on engaging patient advocates in diabetes, we were dismayed to hear examples of the social injustices experienced by those living with diabetes. Mr. Laurent Doyen (French Diabetes Federation), who has lived with type 1 diabetes for 44 years, described how certain jobs are barred to people with diabetes. He also explained that insurance companies charge extra premiums for those with diabetes, noting that he has had to pay over three-times what his partner paid when purchasing insurance to buy a house. Shockingly, discrimination even extends to the healthcare system – Mr. Doyen commented: “you can’t imagine the discrimination [of people with diabetes] in some hospitals.” On November 12, the French Diabetes Federation will submit its recommendations to improve the health system, including stipulations that may address these social injustices. In her closing remarks, IDF Europe Chair-Elect Dr. Niti Pall noted that Mr. Doyen’s description of the social injustices “really touched me.” We were similarly affected and eagerly await the outcomes of the Federation’s proposal.

• We were pleased to see the symposium include a critical panel surrounding the importance of language in diabetes. As moderator Dr. Partha Kar (Portsmouth Hospitals NHS Trust, UK) pointed out, understanding patient preferences for diabetes language is “probably better education” for healthcare professionals than hearing the latest scientific advancements. Indeed, newly elected President of the Swedish Diabetes Association Ms. Cajsa Lindberg shared that when she was first diagnosed with type 1 diabetes as a 13-year-old, a nurse in the hospital instructed her not to prick her thumb or forefinger because she would need them when she became blind – “not ‘if’, but ‘when.’” Years later, Ms. Lindberg still does not prick her thumb or forefinger. To this end, Dr. Jose Manuel Boavida (Portuguese Diabetes Association) emphasized that HCP training must be improved, claiming that diabetes language should be considered equally as important as the physiology of the disease. Ms. Lindberg agreed, noting that she is especially frustrated when HCPs tell newly diagnosed patients that they can live a “normal life.” As she put it, people with diabetes can live “good, heathy, long lives, but it will never be normal. Saying that diminishes all the hard work that goes into diabetes.” Still, physicians are not the only offenders – Ms. Lindberg claimed that “some of the toughest things are said from our own community,” especially regarding the distinction between type 1 diabetes and type 2 diabetes. Often, she explained, language can be especially harmful when people try to oversimplify diabetes.

  • Ms. Valentyna Ocheretenko (member of the Ukrainian Diabetic Federation and former IDF Vice President, 2003-2009) emphasized that not only must healthcare providers take care in how they communicate with patients, but also the degree to which they are understood. She referenced a 2017 University of Missouri study, which found that low health literacy was a significant contributing factor of readmission for chronic conditions such as diabetes, heart disease and cancer; if sufficiently improved, the healthcare system could save $105 billion - $175 billion/year.

  • The final panel discussed two bright spots from the Finnish Diabetes Association and the Turkish Diabetes Foundation. Ms. Sari Koski (Finnish Diabetes Association) lauded the Finnish healthcare system as working “quite well” for people with diabetes, noting a particular emphasis on prevention and early diagnosis of type 2 diabetes. She explained that 20 years ago, Finland had an estimated 250,000 undiagnosed people with diabetes; that number is now ~50,000, even with the number of new diagnoses declining annually. Unfortunately, Ms. Koski didn’t go into the specifics of how this improvement was achieved – we think healthcare systems around the world (especially in the US) would be very interested to know! Indeed, Dr. Pall mentioned that she uses the Finnish system of care and referral pathways as a model for her work helping the Philippines develop their own healthcare system. Professor Meltem Kuruyaziki (Turkish Diabetes Foundation) highlighted several ongoing projects in her country. We were quite impressed by advocacy efforts that culminated in the government providing free treatment for all patients under 18 years-old. Professor Kuruyaziki also described a nationwide, peer diabetes education program that has reached 34,000 patients. After just one year, participants’ A1c dropped from 7.6% to 6.9%, underscoring the immense value of patient education. We wonder if the fact that education was provided by peers was important here, or whether education alone drove the positive results.

Oral Presentations: Diabetes: Gazing into the Crystal Ball

Excess Mortality and Cardiovascular Disease in Type 1 Diabetes in Relation to Age at Disease Onset: A Study of 27,195 Patients with Diabetes

Naveed Sattar, MD, PhD (University of Glasgow, UK)

Dr. Naveed Sattar presented hotly-debated data from his group’s recent publication in The Lancet, indicated that there is an inverse association between type 1 age of diagnosis and excess CV event risk. Previous work has demonstrated that a younger age of onset with type 2 diabetes is more pathogenic in terms of years of life lost for the individual. From this starting point, Dr. Sattar and his colleagues in Sweden set out to answer the question of whether this same trend also applies to individuals with type 1 diabetes. Using the Swedish National Diabetes Register, 27,195 individuals with type 1 diabetes were each matched with five controls (based on of age, sex, and county) and analyzed in comparison for years of life lost and excess risk of mortality and CV disease. Across the spectrum of CV mortality endpoints, risk was unsurprisingly imbalanced toward those with type 1 diabetes; moreover, relative risk increased with an earlier age of onset. This trend was most prominent when examining risk of coronary heart disease, acute myocardial infarction, and cardiovascular disease (see figure below for full results). Fitting this data into a broader research context, Dr. Sattar noted that these findings are “broadly in line” with separate data presented in a 2015 JAMA publication examining the type 1 population in Scotland; this study found an estimated loss of life expectancy, at 20 years old, of approximately 11 years for men and 13 years for women with type 1 when compared to the general population.

• Hypothesizing on a possible mechanism for a link between earlier type 1 onset and excess mortality, Dr. Sattar pointed toward a more aggressive immune phenotype seen with those who develop type 1 earlier. This earlier, more rapid, and more severe loss of beta cell function may explain excess mortality down the line, but Dr. Sattar did not venture too far into these potential mechanistic explanations.

• Not to be alarmist, Dr. Sattar emphasized that the absolute risk of CV events remains low at ~2 CV events per 100,000 patient years in this young cohort. Moreover, for most of the CV endpoints analyzed, the absolute rates were even lower at ~1 case per 100,000 patient years.  That said, once such patients were middle aged, their excess risks of CVD were on average higher than for patients with Type 2 diabetes, emphasizing that overall excess risk in type 1 is higher than most appreciate.

  • Dr. Sattar also noted that the data used in the analysis may be a bit outdated. Older patients in the analysis – who had diabetes for 30+ years – primarily consisted of so-called pre-DCCT individuals. Given very different standards of care before DCCT was conducted, this may have impacted mortality in the cohort and make this sample less representative of patients who are diagnosed with type 1 diabetes today.

• Dr. Sattar advocated for increased use of statins sooner in people with type 1, especially when those who developed the disease at <15 years old reach the age of 30 to 40 years old. Of course, this does not mean that statins should be used for children with type 1 diabetes; rather, that statins should be more aggressively considered and utilization once those with type 1 reach the AHA/ADA recommended 30 to 40 year old benchmark to be considered for such treatment.

Q: The point of reporting the effect of age at diagnosis alone is the major objection that I have. When you have two variables (age at diagnosis and duration) – when you add these variables up, you get age, which is your timescale. You cannot show the effect of any one of these variables alone, you have to show them together some way.

A: We’ve considered your objections and went back and looked and we still find the same thing [same conclusions of increased mortality risk] actually. I’m reasonably comfortable that these statements are broadly correct.

Corporate Symposium: Illuminating the Path of Care for a Multifaceted Disease (Sponsored by Sanofi)

From Banting to Smart Insulins: Advancing Insulin Technology

Richard Bergenstal, MD (International Diabetes Center, Minneapolis, MN)

In a powerhouse opening lecture to Sanofi’s corporate symposium, Dr. Richard Bergenstal highlighted new data on the correlation between time-in-range (TIR) and retinopathy in support of using TIR and time-in-hypoglycemia to govern care decisions and treatment goals. Dr. Bergenstal identified two major headwinds against full acceptance of CGM/TIR, compared to BGM, in clinical trials, regulatory decisions, and even clinical practice. First, TIR currently needs to be contextualized within already-accepted data, endpoints, and standards – that is, A1c – to be accepted; for example, 70% TIR=~7% A1c; ΔTIR 10%=~ΔA1c 0.5%. Second, and giving rise to that issue, is that there is no data on how TIR is associated with complications – “The DCCT was not done with CGM.” This knowledge gap has prompted calls from thought leaders including Dr. Irl Hirsch for a definitive study examining glucose variability and long-term outcomes. However, Dr. Bergenstal emphasized that the needle is starting to move on this point (which we think will also help to move the former), pointing to a just-published Diabetes Care study that demonstrated a significant (p<0.001) correlation between TIR and vision-threatening retinopathy, with those spending <50% TIR experiencing a ~10% risk of the complication (see below). This was a cross-sectional study (n=3,262) examining TIR over 24 hours (via CGM) and fundus photographs, but the association remained when controlling for A1c, age, BMI, diabetes duration, sex, blood pressure, and lipid profile – in our view, quite a compelling result. Moreover, this study joins other analyses supporting the clinical relevance of TIR, including a very convincing analysis of DCCT TIR data and microvascular complications from Dr. Roy Beck, as well as epidemiological data from Australia (cited by Dr. Hirsch) pointing to a reduction in microvascular complications with the introduction of prandial insulin analogs – despite no overall changes in A1c. Dr. Bergenstal stressed that, as the evidence between TIR and vascular complications, particularly the microvascular, builds (including an article that he and Dr. Beck are working on, to be published soon), “TIR and time-in-hypoglycemia will be the new metrics by which we measure diabetes therapies.”

• As Dr. Bergenstal sees it, TIR is a much more rigorous measure of glycemic control compared to A1c – and also more useful, as it can bring patients to “A-HA moments” in which they see first-hand the effects of not only therapeutic agents, but also lifestyle changes, on their current glucose state. He referenced differing A1c goals – AACE recommends ≤6.5%, ADA <7%, and ACP a highly-debated 7%-8% – and pointed to a very recent Diabetes Care study on the legacy effect demonstrating each target in the first year following type 2 diabetes diagnosis affects microvascular complications. Referencing the AACE guideline of ≤6.5%, the study found a 20% increased risk (HR=1.204, 95% CI: 1.063-1.365, p=0.004) of microvascular complications among those with an A1c between 6.5%-7.0% (ADA guidelines) vs. ≤6.5%, a 39% greater risk (HR=1.391, 95% CI: 1.226-1.578, p<0.0001) among those with an A1c between 7.0%-8.0% (ACP guidelines). Moreover, there was a 60% greater risk (HR=1.603, 95% CI: 1.340-1.917, p<0.0001) among those with an A1c between 8.0%-9.0% in the first year after diagnosis, and a highly-concerning >doubled risk (HR=2.213, 95% CI: 1.892-2.590, p<0.0001) for those with an A1c >9.0% (see the slide below). Of course, this was an observational cohort study, so there is some risk that A1c in the year following diagnosis is more of a marker for overall health and risk, but the consistency and magnitude of the trend certainly deserves attention. However, Dr. Bergenstal went on to argue, A1c probably isn’t the best metric to control or examine in the first place. In his words, “If we can’t even agree what the best A1c is, perhaps it’s time to move on and look at the glucose itself through CGM.” Of course, work still has to be done to define time-in-range goals, but our understanding is very much that TIR (and other CGM-based measurements) can offer a more objective, actionable, and precise picture of glucose control.

• In reference to CGM’s many benefits beyond cardiovascular outcomes, Dr. Bergenstal highlighted the HypoDE study, which provided excellent evidence supporting CGM’s efficacy in patients using MDI with hypoglycemia unawareness; results were originally presented at ATTD 2018 and simultaneously published in The Lancet. In the six-month, 12-center trial, CGM (Dexcom G5) significantly reduced hypoglycemia incidence (≤54 mg/dl for at least 20 minutes) by 72%. Severe hypoglycemia was nearly halved with CGM, a major win for cost-effectiveness. These data certainly bolster the case for CGM as standard of care in people with hypoglycemia unawareness, including type 2 – the evidence for the use of CGM in this population received a “C” level grade in the 2018 ADA Standards of Care (published December 2017), and we certainly hope this will change in the next edition! Dr. Bergenstal made a point to advocate for CGM use far beyond just those with hypoglycemia, though, stating, “I don’t see any reason why every type 1 shouldn’t be offered CGM. Not everyone will take that, but everyone should be offered.”

• Much to our excitement, Dr. Bergenstal set out to motivate his audience with a recording from our very own Adam Brown’s new audiobook version of Bright Spots and Landmines. Dr. Bergenstal elected to use the following quote (check out the video here), “We have too much darkness in diabetes. Negativity, confusion, frustration, exhaustion, blame, guilt, and fear. For those of us living with diabetes and the people we love, the cost of this darkness is high. We often don’t know what to do, aren’t doing what we should be doing, feel bad about what we are doing, or are told we’re getting it wrong. We can do so much better.” Dr. Bergenstal then contextualized this quote by referencing to how far insulin has come, “I think Adam said it well. You heard the word fear. You got the sense of the burden of this disease. We can do better […] and I hope you come away with the sense that we are making some progress.” Indeed, Dr. Bergenstal focused heavily on the idea of lowering the burden that diabetes places on patients’ lives, in multiple dimensions, and the role that time-in-range has to play in that goal.

Keynote Lecture: Diabetes: A Heterogeneous and Complex Disease

James Gavin, MD, PhD (Emory University, Atlanta, GA)

In a riveting keynote lecture on the diversity and heterogeneity of diabetes, Emory University’s Dr. James Gavin advocated for a novel classification system to “illuminate the pathway to comprehensive, better-targeted, and more effective treatment of diabetes.”

• Dr. Gavin criticized how, historically, diabetes has been classified into the narrow “treatment boxes” of type 1 (i.e. antibodies and beta cell destruction) vs. type 2 (i.e. insulin resistance). In his view, the beta cell destruction vs. insulin resistance framework is an oversimplification. He described how antibody positive patients can present as if they have type 2 diabetes, and how type 1 patients can present with insulin resistance (i.e., looking like type 2 diabetes). To parallel the oversimplified classification scheme for diabetes, Dr. Gavin argued that the standard framework for diabetes treatment is equally narrow: Insulin for type 1 diabetes and stepwise addition of oral anti-diabetic agents and eventually insulin for type 2 diabetes. He noted that although diabetes diagnoses may have extensive clinical overlap, differences in etiology, phenotype, and (likely) genotype are important further considerations to determine a more individualized treatment plan.

• Dr. Gavin pointed out that the current glucose-centric diagnostic model for diabetes often “handcuffs” clinicians into solely focusing on reducing A1c without adequately addressing its causes or engaging with opportunities to attenuate comorbidities. To this end, he highlighted a new classification method that describes five subtypes within type 2 diabetes, and a beta cell centric model (the “egregious eleven”) which allows for tailored treatment of 11 specific pathways associated with diabetes at both the physiologic and genetic levels.

  • Dr. Gavin pointed to a 2018 Lancet study which created five “clusters” within type 2 diabetes: (1) Severe autoimmune diabetes (SAID); (2) Severe insulin-deficient diabetes (SIDD); (3) Severe insulin-resistant diabetes (SIRD); (4) Mild obesity-related diabetes (MOD); and (5) mild age-related diabetes (MARD). These clusters are based on six clinical variables: (1) Glutamate decarboxylase antibodies (GADA); (2) age at diagnosis; (3) BMI; (4) A1c; and homoeostatic model assessment estimates of (5) beta cell function and (6) insulin resistance. Lending credence to the idea of distinct subtypes within type 2 diabetes is the observation that these cohorts experience very different rates of progression of diabetes-associated complications (see below). Furthermore, there are genetic differences between the cohorts. For instance, a variant in the TCF7L2 gene (previously thought to be associated with type 2 diabetes in general) was associated with SIDD, MOD, and MARD but not SIRD. Importantly, carriers of this TCF7L2 risk-variant do not respond to incretin therapy (e.g., GLP-1 agonists), so SGLT-2 inhibitors or insulin are a better option for A1c reduction.  As such, there is significant potential to leverage the heterogeneity of type 2 diabetes into more individualized, effective therapies (we’re nervous to say “precision medicine” as this has become a buzzword, though this is certainly a step closer to that coveted level of personalization). Dr. Gavin pointed out that not only would this confer better patient outcomes, but would also help avoid unnecessary patient/provider frustration over a less-than-ideal response to an initial prescription (e.g. a GLP-1 agonist for a TCF7L2 risk-variant-carrying patient). This frustration is all too common with the current treat-to-fail paradigm of diabetes prescriptions.

• Dr. Gavin argued that diabetes is even more nuanced than the famous “ominous octet” would suggest, describing an alternative “egregious eleven” which includes the influence of the microbiome, immune system, and increased glucose absorption in the stomach and small intestine in addition to the traditional players – impaired insulin secretion in the beta cells, increased glucagon secretion in the alpha cells, increased hepatic glucose production, neurotransmitter dysfunction, decreased muscular glucose uptake, increased glucose reabsorption in the kidneys, increased lipolysis in adipose tissue, and a decreased incretin effect in the gut. He argued that factors within the egregious eleven are selectively treatable (or likely to be treatable in the future), making possible a targeted and individually-tailored diabetes treatment plan, depending on a patient’s particular pathophysiology. Taking this one step further, Dr. Gavin emphasized that well-designed, early combination regimens have the potential to address most defects concomitantly – a major win for patients and vote of confidence in up-and-coming dual- and tri-agonists, such as GLP-1/GIP and GLP-1/glucagon agonists, that target multiple aspects of the egregious eleven.

• To conclude, Dr. Gavin touched on the value of real-world evidence to complement and expand upon existing RCTs. He argued that management of a disease as complex and heterogeneous as diabetes depends on many factors that we don’t (or can’t) measure in RCTs. RWE therefore represents perhaps the only way we can get insight into critical issues such as patient preferences, likelihood of adherence, impact of cost, and complexity of certain treatment regimens.

Corporate Symposium: From Science to Reality: Understanding the Silent Side of Severe Hypoglycemia (Sponsored by Lilly)

Pratik Choudhary, MD (King’s College London, UK), Thomas Danne, MD (University of Hannover, Germany), Brian Frier, MD (King’s College London, UK)

The message from a star-studded Lilly symposium on day #1 were crystal clear: Hypoglycemia, specifically severe hypoglycemia, remains a tremendous problem in diabetes today, our strategies for managing it are not yet good enough (though speakers pointed out that technology like CGM is helping), and despite these facts, we stigmatize and marginalize conversation about it. Following a discussion that also included hope for the use of technology and novel agents in hypoglycemia management, addressing hypoglycemia in the elderly, and the Emergency Medical Service (EMS) take on severe hypoglycemia, VU Medical Center’s Prof. Frank Snoek summed up sentiments bubbling up under each preceding talk: “We need to start normalizing hypoglycemia and the emotions around it. We need to show we understand it’s a difficult balancing between hyperglycemia and hypoglycemia, and if it occurs, it can be really frightening, it can induce avoiding behaviors, and we need to get the shame and the blame out of the way. We should monitor hypoglycemia as a routine part of consultation – it should be there all of the time.”