Greetings from the City of Angels, where we are reporting live from this year’s AACE meeting. Day #1 was filled with pre-conference proceedings, so we were treated to overviews of all things diabetes technology, as well as diabetes and CVD. See below for our top highlights and check out our conference preview – tomorrow kicks off with a 6:30 AM session on CGM “trends and efficiencies”!

Diabetes Technology Highlights

1. Dr. Bode’s Pump Overview: Insulet Dash Pharmacy Launch Underway, MiniMed 670G Case Studies, Valeritas V-Go Dosing; Estimates 550k US Pumpers

Atlanta Diabetes Associates’ Dr. Bruce Bode gave a whirlwind overview of pump therapy, including a handful of 670G case studies, commentary on Valeritas’ V-Go type 2 insulin delivery patch device, and news that Insulet’s next-gen Omnipod Dash US pharmacy launch is underway. He also ballparked the number of US pumpers at 550,000, though didn’t cite a source. His main take-home messages were that (i) “all of this [automated insulin delivery] works, it doesn’t matter where you are”; and (ii) CGM will be the standard of care.

• Dr. Bode said that Insulet’s new Bluetooth Insulet Dash Omnipod and PDM has now launched in US pharmacies with no upfront cost. We later confirmed with Insulet that this is indeed the case, though we haven’t heard a public announcement along these lines. As of the 4Q18 call, 120 million lives were covered in the pharmacy. We’ll look to hear more from Medical Director Dr. Trang Ly at the corporate symposium tomorrow morning and in the exhibit hall.

• As at ATTD, Dr. Bode compared CareLink data from the ~69,000 US 670G users to those from his own Atlanta-based clinic (n=211). He noted that his patients are doing slightly better in the time-in-range department – 73% vs. 70% – than the broader 670G user population, possibly because his clinic has patients upload data to CareLink weekly until they reach goal so that they can receive guidance from medical staff. He also recapped CareLink data from 60,394 type 1s (69% time-in-range, 75% in auto mode) and 2,443 type 2s (74% time-in-range, 76% in auto mode), and broke out the data by age group, pointing out that those ≥65 years-old (n=3,594) achieved the highest time-in-range (74%) and the greatest time spent in auto mode (80%). “Medicare doesn’t cover this, but they have a lot of data supporting its use.”

  • According to Dr. Bode, the FLAIR study of the next-gen 780G hybrid closed loop (with DreaMed fuzzy logic + auto correction boluses) “in the next month or few months” (ClinicalTrials.gov forecasts a May 13 start) and personalized closed loop “sometime next year.” See a granular rundown of Medtronic’s closed loop pipeline from ATTD here – 780G was slated for “1 year” and the personalized closed loop was slated for “2+ years.” Dr. Bode hopes the following improvements to future systems will begin to “improve compliance”: (i) minimal calibrations of CGM (Tandem’s Control-IQ with G6 is in position to be the first calibration-free hybrid closed loop later this year); (ii) auto-correction with missed meals (e.g., Control-IQ and 780G); (iii) variable glycemic targets (“If you’re pregnant and want a target of 90 mg/dl, that should be possible”); (iv) minimize the guardrails for safety (e.g., minimum basal, maximum basal); (v) sharing the pump/CGM screen with others via Bluetooth (expected from Medtronic within one year); (vi) auto adjustment of carb ratios; and (vii) use of ultra-fast acting insulin (per Dr. Bode at ATTD, Fiasp is currently under investigation in two closed loop studies, including one using Medtronic’s 670G – phase 4, n=20, expected completion May 2019 – and another using Beta Bionics’ iLet system – phase 2, n=24, expected completion April 2019).

  • Dr. Bode presented a handful of 670G case studies – here are a few of the highlights:

    • A 15-year-old type 1 got kicked out of auto mode frequently after missing lunchtime boluses;

    • A 63-year-old type 1 (also an RN CDE) saw frequent nighttime glucose spikes due to high-fat, high-carb snacks like pizza and ice cream before bed. She said she would typically take a square wave bolus for these bedtime snacks, but since that’s not possible with 670G, Dr. Bode recommends a second, smaller bolus later on to handle the snack;

    • A concierge nurse suggested that an 86-year-old type 1 move into a nursing home due to recurrent severe lows. Dr. Bode recommended that she instead go on 670G, and despite severe macular edema (her husband is able to help her navigate the sysem), she spent 99% of time in auto mode and had mean glucose of 157 mg/dl with a standard deviation of 44 mg/dl;

    • A 66-year-old type 2 spent 77% in auto mode and 74% in-range on 670G and is “doing remarkably well.”

• Dr. Bode noted that, contrary to the belief of many endos, the 40 unit/day preset basal (highest dose) V-Go supplies sufficient insulin for most type 2s: “You put people on the 40 unit/day version, and they end up getting a lot of hypoglycemia. What happened? It turns out people tell you they take a lot of insulin, but they often don’t take any.” He added that, in his experience, V-Go users don’t know their preset basal setting, nor the amount of prandial insulin they dosed – however, they are typically able to report how many 2-unit “clicks” they give at each meal/snack. The Bluetooth connected V-Go SIM (Simple Insulin Management) should launch in the US by the end of 2019 (per Valeritas 4Q18), eliminating the need for self-reported insulin dosing histories.

  • Dr. Bode also mentioned that a trial evaluating regular human insulin in V-Go is now underway.

Quotable Quotes

• “It was a big mistake that ADA said give 15 grams of anything to treat lows – they’re eating brownies and anything else. It should really just be gels or tabs.”

• “It’s amazing how many people will wear a flash CGM but they never flash it. That’s an issue. You’d do better with the professional version in that case.”

• “There’s a 1%-3% risk of DKA/year with pumps, which is less than MDI, but it’s still there. And it’s totally preventable.”

• “Whenever the bolus percent is less than 50%, you have to ask if the patient is on a low-carb diet – if not, you are really over-basalized and need more bolus.”

Questions and Answers

Q: Is there more rapid insulin in development for closed loop?

Dr. Bode: Currently, Fiasp has not gone into closed loop yet. In Onset 5 (Fiasp vs. Novolog in non-closed loop pump), the two insulins were equivalent, Fiasp was not superior. Earlier, we were doing 670G with URLi, Lilly’s ultra-rapid acting insulin. The key with these faster acting insulins is we need to make sure they’re stable in a pump. Right now, they’re only stable in syringe-based pumps, not in a pump like Tandem’s, which’ll tear it up. Tandem’s t:sport is syringe-based, so it’ll work.

Dr. Bergenstal: There are people who use inhaled insulin with AID systems. The problem is it doesn’t integrate the numbers in, but it works really well.

Dr. Bode: Mannkind is by themselves, but there’s no question, there’s no question inhaled insulin is very effective.

Adam Brown: But we don’t have to wait for faster insulins to meaningfully move people with the insulins we have today. In FreeStyle Libre real-world data, the time-in-range for those who scanned 10 times per day – the median amount of scanning – was 56%, and they still had four hours over 240 mg/dl. With current insulins, it’s hard to be >240 mg/dl for a long time with 670G. The other thing, even without rapid-acting insulins, if you’re not eating 12 hours a day, basal is good enough to keep people in range.

Q: My tech-savvy millennial patient who is flat, narrow, in-range on hybrid closed loop says, “Why do I still need you?” How do I answer that question? And during pregnancy, when you’re interpreting CGM, is lag time different?

Dr. Bode: There are currently no data with CGM looking at accuracy in pregnancy. You just treat as you normally do, always have them prick finger just in case. We try to get everyone with pregnancy on CGM. [Editor’s note: There is data on CGM in pregnancy, e.g., a 2018 DT&T study of FreeStyle Libre.]

Dr. Bergenstal: It’d be nice if people didn’t need us down the road, as long as there’s a way to get reimbursed for monitoring remotely.

Dr. Bode: There’s no question you need a doctor. Maybe not for glucose, but with diabetes, you have a higher risk of CVD, atherosclerosis, you need to look for complications – you need to do all this and can’t ignore it. For managing glucose, we’re doing a terrible job right now as physicians as a whole.

Adam: I had an endocrinologist visit a month ago, and we did it over video. I sent my Dexcom Clarity data to his EMR over email, and it took 15 minutes end to end. The doctor got paid, I’m sure. He did eventually reply to me to say, “You don’t need an endocrinologist.”

2. Dr. Grunberger Laments Difficulties of Implementing Diabetes Technology in the Real World: Cost and Reimbursement Key Barriers; Company-by-Company Pump/CGM Training Reimbursement Landscape

Dr. George Grunberger lamented the barriers to implementing diabetes technology in an endocrinology office: interoperability, upload time, cost, and reimbursement. After a very detailed overview of the available (but declining) reimbursement for providers that is pitched against the backdrop of flat clinic expenses, Dr. Grunberger displayed a meme with a confused cat and simple caption: “I dunno man, I’m so confused.” He reviewed barriers to technology implementation at a granular level:

• The process of uploading can take anywhere from two minutes for glucose data to 10+ minutes for pumps and CGMs – all of which must be done before the provider sees the patient since they only have 5-7 minutes together on average, four times per year (Dr. Grunberger’s numbers). Fortunately, app-based device systems are making their way on the market, allowing data to be passively sent to the cloud and visualized in clinic without a download.

• Proprietary cables and protocols may be required for various devices (though platforms like Tidepool and Glooko/Diasend are beginning to ease these issues), also necessitating training of staff on each hardware and software. For example, Dr. Grunberger’s clinic staff currently must master at least 15 different diabetes management software systems in order to accommodate all of the devices that patients bring in. We’re not sure why the clinic hasn’t fully transitioned to Tidepool or Glooko for pan-diabetes management needs.

• According to Dr. Grunberger, there is low patient demand for technology (though CGM is in large part driving diabetes industry growth), and few are willing to download their data at home: “For every single patient who uses any device in our clinic, we give them a physical piece of paper that says, ‘please download your data at home.’ You know how many patients have followed that advice in 20 years? Maybe three.”

• Reimbursement (when possible) is poor. Three common codes for CGM onboarding (95249, 95250, and 95251), which cover initial set up of CGM (personal and professional) in a face-to-face encounter and CGM data interpretation are woefully under-reimbursed in Dr. Grunberger’s practice, with average rates near and below 50% of the billed amount (see below). This comes in tandem with the fact that these reimbursement rates are lower today than they have been historically, and amounts to 50% of CGM starts and interpretation essentially amounting to charity work.

Taking one step back, technology is a non-starter without space, staff, hardware, and software, all of which require money, time, and energy, all of which could potentially be better used for other quality improvement initiatives, causing many clinicians to question whether an investment in technology is worthwhile. Additionally, technology may require providers to be on 365/24/7. For example, Dr. Grunberger briefly mentioned that one person with diabetes (not one his patients) who knew one of his patients once called him at 10 pm on a Saturday night due to a pump failure while she was on a camping trip in Europe without another means to administer insulin. Taken together, the startup costs of technology and change in practices they require can appear to be insurmountable barriers for some providers. However, Dr. Grunberger asked the very important question, “Who will take care of the ever-growing diabetes population?” Technology, even if complicated at the outset, can aggregate data into quick and actionable insights, enable between-site “check-ins” via remote care, and promote lifestyle changes through meaningful feedback. While he mentioned a concern that these technologies could potentially replace endocrinologists altogether, we think the more likely situation is that they reduce burden and improve workflow, so that endocrinologists can spend more time with those who need the attention most.

• Given that no billing codes for pump training yet exist (and CGM codes are under-reimbursed), Dr. Grunberger listed the manufacturer-provided compensation for pump/CGM education. In his words, “thank God for competition” for driving down costs, though “[providers] certainly won’t get rich off of these training reimbursements.”

  • Medtronic, which he picked as having “the best CPT education, follow-up, and trouble-shooting” reimburses $300 for new pump training, $200 for training on pump upgrades to new platforms, and $75/hour for training on the same platform. The company also provides $300 for new CGM sensor training, $100 per upgrade and $75 per hour for additional support up to two hours.

  • Tandem reimburses $300 for a new pump start, $250 for a switch from a competitor, and $200 per patient for group training, but nothing for sensor (Dexcom G6) training.

  • Insulet provides $300 for an individual patient start and training, $350 for a group of two, $400 for a group of three, and $200 for a switch from a competitor.

  • Dexcom does not provide 1:1 training (only phone or Skype) and does not pay practices to do this. To get around this, Dr. Grunberger’s clinic schedules the patient with a nurse, arranges a brief clinician encounter, and bills 95214 or 95215 if it’s a prolonged service (both for pre-CGM evaluations).

  • Abbott also does not provide 1:1 training and does not pay practices to do this, so Dr. Grunberger’s clinic follows the same workflow as with Dexcom.

  • Senseonics’ is tricky since the Eversense CGM requires a procedure, and no procedural codes yet exist to start or implement diabetes devices. For this reason, the company has provided step-by-step billing codes for providers – see below. However, the long-term CGM CPT codes seen in steps two and four are new and “payers don’t like to pay.” Dr. Grunberger is frustrated that, for example, many of the BCBS’s cover Eversense, but BCBS Michigan does not. Even when they do pay, the reimbursement spectrum is incredibly broad, ranging from $80 to $450 for the first procedure, including system activation and patient training (CPT 0446T) and $600 to $800 for removal of the first sensor and insertion of a second sensor at a new site (CPT 0448T) – see below. This lack of standardization is quite alarming – the same procedure can one provider 5.6x more than another!

3. Adam Brown: Diabetes Apps Can Make Our Lives Easier, but Still Early Days; Useful Apps Are Radically Convenient, Frictionless and/or Present Compelling Decision Support

Our own Mr. Adam Brown presented on the state of diabetes apps, largely reiterating his ADA 2018 talk – see the summary from that talk here, download today’s slides here, and read on for some of today’s highlights. As at ADA, the talk aligned on two key features Adam believes will characterize truly useful apps: (i) radically convenient, frictionless tools; and (ii) compelling decision support. He arrived at these characteristics by asking himself what it is that makes non-diabetes apps, such as Google Maps, Uber, and Spotify, useful. Adam then proceeded to call out 40 different examples of diabetes apps – either currently out or in development – with exciting prospects (see a list of them here, along with an explanation of why they are useful).

• dQ&A panel data on the weekly use of apps to manage diabetes indicated no significant changes from 1Q18 to 3Q18: In Q3, 12% of type 2s (n=3,197), 29% of type 1 adults (n=1,663), and 54% of type 1 kids’ parents (n=153) used apps – no marked changes from 1Q18. App adoption remains astoundingly low, despite the fact that nearly everyone in dQ&A’s panel owns a smartphone and many popular diabetes apps are free.

• Adam’s discussion of radically convenient and frictionless tools touched on:

  • Display apps that allow people to carry fewer devices around with them and/or pull their dedicated diabetes devices out less;

  • Data management apps that eliminate the archetypal “spaghetti mess of wires” in a clinic – he mentioned that in the Jaeb/Tidepool Loop observational study in which he is participating, Tidepool is remotely aggregating all of his blood glucose data and he mailed a blood sample through the mail for A1c testing (he hasn’t visited a clinic once, “a total gamechanger”);

  • Mobile CGM apps that allow users to view their data from any given day in the past, as well as those that push insights on blood glucose trends to the user (either in the moment or retrospective);

  • Connected glucose meters only have 13% uptake in the dQ&A panel. In Adam’s view, this is in part because there is not yet enough compelling decision support to make connectivity a must-have;

  • Subscription “connected care” companies (e.g., One Drop, mySugr) that are frictionless in that they automatically ship supplies to the user as needed and provide touch points between clinic visits – Adam suggested that these players could do a better job of sharing outcomes and integrating with the healthcare system. We do note progress on both of these fronts.

• His discussion of compelling decision support was centered on the idea that the goal of apps is to reduce burden and “take some of the 42 factors that affect blood glucose off the table.” Promising areas he touched on include:

  • Welldoc’s BlueStar type 2 diabetes management app, which offers real-time, in-app coaching. He emphasized that glucose monitors and meters are just tools, not therapies, and in-app prompts, messages, and coaching help people extract more value from glucose data. He also noted that Welldoc’s product for type 1 diabetes and CGM integration are currently under FDA review;

  • Basal insulin titration for type 2 diabetes, which he called the “most perplexing area of diabetes technology,” given its no-brainer potential yet slow commercialization. There are currently at least five FDA-cleared basal titration apps, yet relatively little commercial traction, but Adam thinks we’ll see more action in this sphere, especially as pharmaceutical companies delve deeper into smart pens.

    • He also mentioned bolus calculators and DreaMed/Glooko’s FDA-cleared Advisor Pro as other “obvious areas” of decision support, then alluded to CGM-based bolus calculators which will likely lend to smarter dosing advice that takes blood glucose trends into account.

  • Closing the loop for injection users via smart pens, CGM, and dosing advice on an app (á la Lilly, Bigfoot, Dexcom, etc.). He estimated that this area is ~3 years behind where closing the loop is today for pumps, “but it’ll move fast because most people aren’t on pumps.”

  • App-based automated insulin delivery. Adam shared his two years of experience on DIY closed loop, depicting improved time-in-range (73%->81%; an improvement of one entire month per year!) and the insanity of going to bed with the same basal rate every night. This latter point was illustrated by two separate nights within the same week: One night, his hybrid closed loop system needed to deliver 16 units in order for him to wake up in range, and on the other, it only needed to deliver 6 units. “This is why it’s ludicrous in type 1 diabetes to go to bed every night with the same basal rate, unless you’re a ridiculously regimented person.”

• Adam closed with the Home Screen Test: “Is a diabetes app so useful it makes it on to a phone’s home screen? Can it make it down to the bottom dock?” The Dexcom CGM and the DIY Loop apps have both made it to Adam’s bottom dock because they provide radical convenience (no fingersticks, no carrying stuff around) and compelling decision support (automated insulin delivery).

Questions and Answers

Q: Which app is a good starting point for my patients to get their feet wet in the app world?

Adam: I think mySugr and One Drop are good starting points. One Drop also has diabetes education courses within their app, which could be really useful.

Dr. Bruce Bode: Obviously, many apps are used for patients, but we need to be able to download and look at everything in the clinic too. Tidepool is probably the easiest thing if you want to use only one, since they also cover Medtronic…The one way you can be very successful is you can bill for CGM. The last three years is the only time we’ve seen traction there. Before, we were losing $150,000 per year; we gained $120,000 this year, and I gave it back to educators. It was a significant amount, the first time they ever got a bonus.

Q: Can you comment on how tech and apps may particularly impact care for socioeconomically-challenged type 2s and type 1s in high income, and then later low- and middle-income countries? In Kenya, for example, 90%-95% of people have smartphones.

Dr. Bode: Obviously, with lower socioeconomic status, people need to have the wherewithal to get devices, pumps, and even insulin. CMS is covering CGM, and will cover hybrid closed loop soon, with Tandem, for sure, once it’s nonadjunctive. Medicaid covers it. Going out to Kenya and all, everything has to be simple, plug and play, and without much hassle – it’ll be there.

Dr. Bergenstal: We have not done a good job here. We’re getting worse despite more technology. I’m convinced we’re moving people with 7.1% A1cs down to 6.7% – we need to do a better job. That’ll take healthcare reform as well.

Q: Subscription apps – what is the cost on those? Any reimbursement from insurance?

Adam: Most of the companies are direct-to-consumer without insurance. For unlimited strips, that’ll cost about ~$50 per month. Or less than that. The pros are that there is no prescription and no insurance, but the con is that it might be out of reach for some still.

Dr. Bode: Others like Onduo and One Drop are done by insurance companies. That’s totally free, they don’t pay anything for that.

4. Drs. Bergenstal and Hirsch Break Down the AGP: GMI and A1c Discrepancies to Inform Decision-making, Differences Between Personal and Population Times-in-Ranges Displays, & Case Study Pearls

Drs. Richard Bergenstal and Irl Hirsch presented “GMI” (glucose management indicator), the newly-agreed-on-term for estimated A1c based on average blood glucose, as a means to inform interpretation of A1c, giving concrete examples where the lab-value A1c is “fake news” compared to the mean glucose values. One patient with iron deficiency had an A1c of 8.2% and a GMI of just 6.6%, presenting a potentially dangerous situation if more aggressive treatment was undertaken. Another with stage 4 CKD had an A1c of 8.3% when starting glimepiride. Nine months later, A1c had dropped to 7.7%, but she was spending ~6 hours per day below 70 mg/dL, conferring a GMI of 5.7% (see below). Dr. Hirsch halted the glimepiride, and the patient’s GMI returned to 6.4% with just 1% of time spent below 70 mg/dL – a perfect example of how A1c can be dangerously misleading. At a population level, Dr. Jordan Perlman, a fellow at UW, presented data at ADA 2018 showing that ~33% of people with diabetes have A1cs that can be classified as not aligning with the value expected from CGM/SMBG. The takeaway was perhaps best stated by Dr. Bergenstal at DTM 2018: “If your A1c in lab is 7% but your GMI is always saying 6.5% and derived from a reliable 14 days or more of CGM, then you probably need to think about your targets of A1c to be safe. Maybe you need to run a little higher because for your personal glucose values, you’ll be running a little lower than others with the same A1c.”

• Dr. Bergenstal cleared the air around the intended use of percent time-in-ranges that add up to 100% and those that exceed 100%. As seen in the image below, the former splits time in hypoglycemia into 54-69 mg/dl and <54 mg/dl (i.e., all percentages add up to 100), while the latter double counts the time <54 mg/dl in the next category up, <70 mg/dl (i.e., all percentages add up to more than 100). Dr. Bergenstal explained that the version that adds up to 100% is intended for personal/clinical use, while the version that double counts level 2 hypoglycemia and hyperglycemia is more useful for populations/research. This distinction is an important methodological caveat that should be watched for in studies.

• As we first heard at IDC 2019, Dr. Bergenstal has expanded his litany of useful mnemonics, using MGLR (more-green less-red) for times-in-ranges and FNIR (flat, narrow and in-range) for CGM tracings as easy heuristics to drive decision-making. These add to his nine steps for interpreting CGM data with a patient – which he underscored as a great tool for onboarding PCPs to new diabetes technology – and three P’s of diabetes care in 2019 (precision, personalized, and patient-centered). Taken together, these lessons provide the broad framework necessary to address the at-first-glance overwhelming amount of data produced by novel diabetes technologies that must be sifted through by patients and providers, often in 15-20 minutes or less.

5. Pearls from Dr. Hirsch’s Case Studies: InPen “Extremely Popular” at His Clinic, Indicators of “Fake Carbs” with 670G, Gross Over-Basalization

A series of case studies presented by Dr. Irl Hirsch brought his personal insights to interpreting insulin pump and glucose sensor data in real world settings. Some key takeaways:

• Companion Medical’s InPen has become “extremely popular” at Dr. Hirsch’s clinic, especially for patients who wish to switch off of a pump. Companion recently announced that its device is now available in retail pharmacies. For some, particularly those who can stay in auto-mode for ≥80% of the time, hybrid-closed-loop (Medtronic’s 670G) can significantly improve glycemic outcomes and quality-of-life. Others, such as those who are overly aggressive with correction doses at bedtime, would benefit from a smart pen (e.g., Companion’s InPen).

• With the 670G, he posited that anyone with a mean glucose value below 140 mg/dl is possibly inputting “fake/phantom carbs” to drive more aggressive treatment. Other indicators of fake carbs on 670G include more than 4% of time spent below 70% mg/dl, high standard deviations, and a high bolus-to-basal ratio.

• “We’ve been grossly, grossly over-basalizing our patients. What we have found is that most people need basal insulin between 40%-45% of their total dose. When I see someone who is struggling and above 50% basal, I know we need to make carb ratios more aggressive to increase the prandial. Micro-boluses will never catch up when a patient gives their insulin after they eat.”

• Providers (and their practices) must be literate in the spectrum of diabetes devices. Dr. Hirsch identified: (i) a lack of infrastructure in offices leading to inefficiency and perceived costs in time and money; and (ii) a lack of training for the lack of data downloading in clinics today. There is no doubt that data downloading has become an integral component of modern-day therapy (and will only grow further), and Dr. Hirsch questioned whether many of the perceived attitudes about time and money even apply today as downloads become passive and interoperable with smartphones.

• GMI should be used in tandem with A1c to determine targets. If the report you receive does not include GMI, then go to http://www.jaeb.org/gmi to quickly calculate it with mean glucose.

• The coefficient of variation (CV) is often used as the marker of glycemic variability – for which Dr. Hirsch believes the target should be <33% – but providers can also multiply the standard deviation by three and compare to the mean glucose if CV is not given. If the product is less than the mean glucose, then the glycemic variability is acceptable. If not, adjustments should be made to reduce glycemic variability.

• Insulin dosing tweaks should be made on a case-by-case basis and aren’t always necessary. “I had an 82-year-old patient taking insulin glargine and aspart with an A1c of 6.8%. She has had type 1 for 55 years with an average mean glucose of 140 mg/dl and CV of 31% – all strong. However, she had a late-night snack spike that stood out on her AGP. You know what I did? Nothing! She’s 82; she won the war. Enjoy the cereal.”

Diabetes Therapy Highlights

1. Preventing and Treating Heart Failure with SGLT-2s: Dr. Handelsman Calls on Clinicians to Bring More Vigilance to One of Diabetes’ Most Morbid Complications

Dr. Yehuda Handelsman urged his audience to be proactive in the prevention of heart failure, explaining the negative outcomes associated with heart failure events and pointing to a preventative role for SGLT-2 inhibitors. He opened by commenting that, even one year ago, he’d never have thought to be giving a talk on the subject of heart failure, reflecting increasing recognition of the relationship between heart failure and diabetes. When it comes to diabetes, heart failure is uniquely problematic: In contrast to MI and stroke, traditional cardiovascular risk factor control does not impact heart failure risk, and CVOT data show increased mortality following heart failure hospitalizations – in the ELIXA CVOT for Sanofi’s Adlyxin, 18% of all deaths occurred post-HHF, with a HR of 9.3 (95% CI: 7.2-11.9) compared to the group without heart failure hospitalizations (of course, HHF identifies higher-risk patients to some degree). This context led Dr. Handelsman to the old adage that prevention is the best medicine, but he explained that SGLT-2 inhibitors are currently the best medicine to achieve prevention. As early as 2016, ESC’s heart failure guidelines actually recommended empagliflozin (Lilly/BI’s Jardiance) for the prevention/delay of heart failure onset in people with type 2 diabetes – a precocious recommendation that Dr. Handelsman understandably characterized as “too early,” commenting that it created the impression that empagliflozin is only good for preventing HF when there are no symptoms. However, with evidence from three SGLT-2 inhibitor CVOTs (EMPA-REG OUTCOME, CANVAS, and DECLARE) and real-world studies in place, ADA/EASD in 2018 recommended SGLT-2s for the treatment of heart failure and CKD – a decision that CREDENCE (renal outcomes with J&J’s Invokana in type 2 and CKD) has since supported. And while treatment is different from prevention, there is growing consensus – strongly rooted in CVOT evidence – that SGLT-2 inhibitors exert a protective effect against heart failure hospitalizations in people with and without known heart failure (see AHA 2018, CODHy 2019); Dr. Handelsman emphasized large primary prevention cohorts in DECLARE and CREDENCE in particular, noting that HF risk reduction extends across those with and without known CVD or HF at baseline. Given there remain no drugs indicated for the treatment of heart failure with preserved ejection fraction (HFpEF) – the type of heart failure thought to overlap most strongly with diabetes and obesity – we would love to see greater utilization of SGLT-2 inhibitors to prevent heart failure onset in the first place, as well as in patients with existing heart failure. To this end, we appreciated Dr. Handelsman’s call to the audience, “We [as endocrinologists] don’t need to manage CHF, but we need to diagnose it early enough or, better yet, prevent it,” highlighting the importance of vigilant monitoring for comorbidities among people with diabetes.

2. Dr. Daniel Einhorn Says Metformin “Likely Cardioprotective,” Citing MET-REMODEL and Anticipating VA-IMPACT; Critiques Continued Widespread Use of DPP-4 Inhibitors

In an engaging talk on hyperglycemia management in the CVOT era, Dr. Daniel Einhorn provided interesting commentary on the role of various therapy classes in current diabetes care. In particular, Dr. Einhorn strongly supported metformin’s continued use as first-line therapy and pointed to forthcoming evidence regarding metformin’s potential cardioprotective effects. Metformin has a highly desirable blend of qualities: it’s efficacious in terms of glucose lowering and (slight) weight loss, has a clean safety profile with low hypoglycemia risk, is highly affordable, and harbors “likely” CV benefits. On this last point, Dr. Einhorn pointed to two studies – one recently completed (MET-REMODEL) and one expected to complete in 2024 (VA-IMPACT). We first heard results from the small MET-REMODEL study (n=68) at ESC 2018, where Dr. Chim Lang presented data indicating that metformin treatment in patients with prediabetes/insulin resistance and coronary artery disease, without hypertension, reduces left ventricular mass (considered one of the best surrogate markers for CV outcomes). On a primary endpoint of left ventricular mass index, metformin gave a significant reduction over 12 months compared to placebo (-2.71 g/m^1.7 vs. -1.34 g/m^1.7, p =0.033). Full results from this study were published last week in European Heart Journal. Dr. Einhorn noted that although the primary endpoint of left ventricular mass is a useful surrogate marker of CV outcomes, conclusive evidence is needed in the form of a large CVOT. Luckily, such a study is underway in the VA-IMPACT trial, an 8,000-participant CVOT of metformin in people with prediabetes and atherosclerotic CVD – so, a very similar population to that of MET-REMODEL. Expected completion is in August 2024, and we expect this VA-sponsored trial to shed valuable light on the role of prediabetes as a CV risk factor, the benefits of earlier intervention in people with impaired glucose tolerance, and the role of metformin in diabetes and CV prevention. Surely, positive results from the small MET-REMODEL trial bode well for the larger VA-IMPACT trial, and we eagerly await the full results. We also note that another metformin CVOT, the UK’s GLINT trial (n=~12,000), is expected to read out in 2024. See our initial coverage of GLINT from back at Diabetes UK 2013.

• Dr. Einhorn is less enthusiastic about the continued use of DPP-4 inhibitors and predicts that usage will decline in the near future. He expressed skepticism concerning the cost/benefit profile of this therapy class, noting that “if I’m going to pay top dollar for a drug, I expect it to be more than just not dangerous and have a little effect on glucose.” Of course, DPP-4 inhibitors are one of the most widely used branded diabetes drugs on the market today and are a fan-favorite amongst both patients and providers because of their nearly squeaky-clean safety profile and ease of use. Nonetheless, Dr. Einhorn’s point stands: In a budding era of diabetes care centered on reducing adverse cardiovascular and renal events in addition to microvascular complications, DPP-4s do not stack up well with GLP-1s and SGLT-2s, as both of these classes boast positive outcomes data that DPP-4s lack from large, randomized trials. Despite resounding neutrality in CVOTs and solid-but-unspectacular glucose lowering abilities, DPP-4s are still quite pricey, and Dr. Einhorn expressed disappointment that these drugs are still “used as much as they are.” Dr. Einhorn is the latest thought leader to express these views on the class: At Keystone 2018, Dr. Ralph DeFronzo characterized DPP-4s negatively, and Drs. Jay Skyler and Steven Nissen have both also articulated similar sentiments. On balance, we believe this class with a very clean side effect and safety profile is just better used in people early in disease progression. Notably, the class will go generic ahead of GLP-1s and SGLT-2s (and, as an oral, will be easier for generic drug makers to manufacture than GLP-1s), ensuring their continued use. Our view is ultimately that DPP-4s still have an important role to play in certain patient groups, particularly compared to SFUs, and we hope to learn what impact this class could have in prediabetes, especially as generic DPP-4s become available in the near future.

3. Dr. Jellinger Dives into Cholesterol Guidelines, Highlighting Benefits of Aggressive LDL-C Lowering (No Lower Limit), Heterogeneity of CV Risk in Diabetes (Use Coronary Calcium), and Critiquing AHA/ACC Recommendations (Too Conservative)

In a guideline-focused talk, Dr. Paul Jellinger emphasized the benefit of aggressive LDL-C lowering and highlighted the lack of a lower limit on the benefit that can be achieved. Looking back on AACE’s 2017 lipid guidelines, he called the establishment of an “extreme risk” category with a <55 mg/dl LDL-C target the most significant contribution of the guidelines to the cardiology, endocrinology, and lipid worlds – indeed, the AHA/ACC’s newest (2018) guidelines have followed suit with their own, brand-new “very high risk” group (though without such a low LDL target). Dr. Jellinger noted that, to date, no other guidelines have recommended a target <55 mg/dl for any patient group, and AACE has applied this target to anyone with (i) progressive ASCVD (including unstable angina) after achieving LDL-C <70 mg/dl, (ii) established CVD and type 2 diabetes, or stage 3 or 4 CKD, or heterozygous familial hypercholesterolemia, and (iii) history of premature ASCVD (<55 years old male, <65 female). Supporting this recommendation, Dr. Jellinger showed that evidence from cholesterol-lowering trials indicates there’s no limit to the benefit garnered by lowering LDL-C, at least in secondary prevention: Statin trials form a straight line for CHD events vs. LDL-C (i.e., lower achieved LDL = lower event rate), predicted to reach “0” around 30 mg/dl. The FOURIER CVOT for Amgen’s PCSK9 inhibitor Repatha supports this finding with very low LDL-C values only possible with PCSK9 use – Dr. Jellinger specifically pointed to a post hoc of prespecified LDL-C groups showing additional benefit all the way down to LDL-C <10 mg/dl, “There is no threshold: The lower you go, the more events you prevent.”

• Stressing heterogeneity within diabetes and the importance of considering CV risk on an individual basis, Dr. Jellinger demonstrated that diabetes is not always a coronary artery disease equivalent. He pointed to MESA (Multi-Ethnic Study of Atherosclerosis), the 10-year outcomes study (n=6,814) that found coronary artery calcification (CAC) independently predicts CVD risk in patients without clinical CVD at baseline – and is the best risk predictor in low-risk patients. Based on this, Dr. Jellinger favors the MESA risk score for incorporating CAC: “We’re in an age where looking at risk factors without calcium is not right.” Importantly, he showed, CAC scoring is as useful in people with diabetes as those without; he explained that the field used to think of diabetes as an automatic CAD equivalent, but MESA found that many people with diabetes scored a “0” on CAC: “Don’t automatically assume all diabetes patients are at ultimate risk.”

• Dr. Jellinger criticized the AHA/ACC’s 2018 cholesterol guidelines for their definition of “very high risk” ASCVD, asserting that the algorithm too severely limits the use of PCSK9 inhibitors to this strictly-defined subset. Per AHA/ACC, “very high risk” includes either (i) history of multiple major ASCVD events or (ii) one major ASCVD event and multiple high risk conditions – he called this a “tall order.” Further, PCSK9s are only recommended after statins and ezetimibe and at LDL-C ≥70 mg/dl. Dr. Jellinger particularly challenged the guidelines’ definition of major ASCVD events to include only (i) recent ACS within 12 months, (ii) history of MI other than recent ACS, (iii) history of ischemic stroke, and (iv) symptomatic PAD (claudication with ABI <0.85 or previous revascularization or amputation). Missing from this list, he said, are (i) CABG (this is listed as a “high-risk condition” instead), (ii) stent placement, and (iii) TIA. On balance, he did praise the inclusion of congestive heart failure as a “high-risk condition” but maintained that these recommendations were too conservative. In contrast, AACE’s “very high risk” category – which has an LDL-C target of <70 mg/dl – includes all patients with type 2 diabetes or stage 3-4 CKD with ≥1 risk factor, as well as anyone with “established or recent hospitalization for ACS, coronary, carotid, or peripheral vascular disease, 10-year risk >20%” or heterozygous familial hypercholesterolemia – a much broader group. On a final note, he challenged the complexity of the AHA/ACC’s primary prevention algorithm and the recommendation to “tone down” treatment at age 70.

-- by Ann Carracher, Martin Kurian, Brian Levine, Peter Rentzepis, and Kelly Close