June 7-11, 2019; San Francisco, CA; Day #1 Highlights

Greetings from San Francisco and day #1 of ADA 2019! This report contains all of our top highlights in diabetes tech and therapy. Remember, you can always check out our ADA 2019 Resource Hub to peruse our Conference Preview or any of our 14 Category Documents – click here for a preview of day #2. A can’t-miss session? Early-morning orals on incretins.

Diabetes Technology Highlights

1. Medtronic and Dexcom sign on to Tidepool Loop, joining Insulet in major move to AID Component Interoperability via iCGM and ACE Pump!

In major automated insulin delivery (AID) and interoperability news, both Medtronic and Dexcom have become official Tidepool Loop partners! Medtronic will integrate with Tidepool Loop via an in-development Bluetooth-enabled ACE Pump and Guardian Sensor 3 iCGM (both yet to be submitted to FDA), and Dexcom will integrate via its already-available, FDA-cleared G6 iCGM. Wow! While Dexcom’s participation was a near-certainty (it’s already being used by many Loopers already using Insulet), Medtronic’s partnership was more of a question mark, although it was clear that many in the diabetes community, broadly speaking, were waiting for this and have been encouraging Medtronic to move toward interoperability for some time. This move represents a bold, welcome, strategic embrace of interoperability and patient choice as key innovation drivers. Tidepool CEO Howard Look excitedly announced the news at the D-Data Exchange today, and Medtronic sent out a corporate press release sharing this news that clearly shows commitment to the program. (While Dexcom did not send out a press release, Tidepool’s blog has excellent posts on both partnerships – here and here). Medtronic and Dexcom join Tidepool Loop’s first partner, Insulet’s Omnipod (November 2018), leaving just Tandem, Abbott, and Senseonics as US players not yet on board. All three could join at some point – Tandem obtained the ACE Pump indication in February, FreeStyle Libre 2 is under FDA review as an iCGM, and Senseonics has plans to submit for iCGM designation. The Loop momentum has been fast: Tidepool announced plans to make a commercial, FDA-regulated version of the DIY Loop app in October – compatible with multiple, in-warranty, Bluetooth-enabled devices, available via the iOS App Store, and eliminating the RileyLink – and eight months later there are now three major players on board. Certainly the new FDA pathways (iCGM, ACE Pump) have been major enablers here. Of note, Dexcom, Insulet, and Medtronic have their own in-house/partnered AID efforts, and integration with Tidepool Loop will be meaningfully additive to those efforts. Tidepool continues to work on an iController de novo algorithm submission to FDA, the third critical component of the mix-and-match AID ecosystem. At DiabetesMine’s annual D-Data today, FDA’s Dr. Alain Silk suggested potential to make the AID controller class II, 510(k)-regulated software with special controls – bringing the whole closed-loop category down from its current class III PMA level, regulating the components separately, and allowing them to be stitched together and updated quickly. The DIY Loop observational study remains underway; we’re not sure when that data will be submitted to the FDA to support approval of Tidepool Loop. If we had to guess, an initial launch of Tidepool Loop with the G6 iCGM seems possible later in 2020, though that is complete speculation and depends on regulatory progress and ACE Pump clearances for Insulet and/or Medtronic.

• Medtronic could not share specific timing on when its ACE Pump or iCGM designations will be filed, but did confirm these submissions will be separate from its planned Bluetooth-enabled MiniMed 780G advanced hybrid closed loop (expected launch by April 2020). Medtronic has not given iCGM FDA submission timing of late, but we assume it will come after the non-adjunctive FDA filing expected soon. We should hear more on Sunday during its ADA Analyst Briefing. We wonder if Medtronic would choose to make a no-screen tubed pump for the Tidepool Loop app, which could reduce the pump’s size and cost. 

• Tidepool characterized the response to this news as “overwhelmingly positive” across social media and via email, with many happy messages and many “when is Tandem joining?” questions. Tandem is certainly the next obvious candidate to join Tidepool Loop, given its status as an ACE Pump with built-in Bluetooth. We assume it is a matter or internal resources and getting the business relationship in place; perhaps Dexcom joining will accelerate Tandem.

• We estimate over 2,500 people are using the DIY Loop iPhone app globally, up from Tidepool’s October estimate of 1,000-1,500. The recent addition of DIY Loop compatibility with Omnipod offers in-warranty Insulet pump users (non-Dash) the ability to Loop, dramatically expanding the funnel beyond old Medtronic pump users. As we understand it, there were already nearly 2,000 users using old Medtronic pumps alone, and nearly 1,500 additional orders have already gone out since the Omnipod compatibility emerged, with new orders coming in at a fast pace. Kelly’s (n=1) been using it for about a month and loves it (“I love reviewing my carb estimates – it shows me when I’ve estimated 15 carbs and I’ve actually absorbed twice as many carbs …”) although she does acknowledge she didn’t set it up; fellow type 1s Amy Tenderich and Cherise Shockley have also expressed great enthusiasm for Omnipod Loop.

• Bringing the Tidepool story full circle, this also means Tidepool Loop users will be able to use a Medtronic pump and Dexcom CGM to automate insulin delivery from an iPhone. (Or any other such iCGM/ACE Pump combination, once the components have received the interoperable clearances, the companies have the business terms in place, and labels have been updated.) Those who have followed Tidepool’s story will remember that combining Medtronic pump and Dexcom CGM data was a major motivation that prompted the initial diabetes data hub; now, these two competitors’ devices will be combinable in a single AID system! 

• Tidepool confirmed with us that Loop will need a prescription. All CGM currently needs a prescription – we hope to see progress and change here since this isn’t the case throughout Europe. Other questions on our mind: When will data be submitted to FDA from the ongoing Jaeb/Tidepool observational study? When will Tidepool Loop obtain FDA clearance/approval – and will it be under a new category of device? Will users be able to modify Tidepool Loop’s glycemic target (and within what bounds)? Will Tandem and Abbott and Senseonics integrate with Tidepool Loop? How will Tidepool support customer service and training? What will Loop’s business model look like? How will users weigh the pros/cons of component AID with Tidepool Loop (mix-and-match companies’ devices) vs. traditional AID (e.g., Medtronic’s MiniMed 670G/780G, Tandem’s Control-IQ, Insulet Horizon). 

2. Dr. Anne Peters: “I Don’t Care if You’re Rich or Poor, I Believe You Need Access to High Quality Diabetes Care;” Developing a DIY Diabetes Survival Kit for Underserved Populations

The inimitable Dr. Anne Peters (USC Keck School of Medicine, Los Angeles, CA) delivered a powerful presentation on the barriers faced by patients living in under-resourced communities. “I don’t care if you’re rich or poor, I believe you need access to high quality diabetes care.” Drawing from her experiences treating patients at her diabetes clinic in East Los Angeles, she outlined two basic tenets for treating underserved patients: (i) technology alone fixes nothing; and (ii) technology combined with “unbelievably” persistent, actualizable, and incremental alterations in treatment can change everything. Dr. Peters rattled off a slew of challenges, including a medical system designed for a different demographic (“there’s a lot of energy around type 1 diabetes and technology that my patients don’t really touch” – pictures of caucasian Apple Watch users on CGM websites don’t necessarily resonate for obvious reasons); complex and stressful life situations (“diabetes may be way over here, while they’re running out of food and facing eviction”); and distrust of the traditional healthcare system. Her ask? Simple solutions with a high benefit that don’t disrupt patients’ already fragile system of existence. As she put it, “they need tools that work right now in the real world.” Sometimes, these solutions can have unintentional consequences. To this end, she shared unpublished data from the STEPP-UP study, an intervention funded by the Helmsley Charitable Trust to develop low-literacy educational tools for diabetes devices. While she was unable to provide the details, “the short story” showed a statistically significant increase in diabetes knowledge, decrease in diabetes distress and depression, and “an extreme increase” in fear of hypoglycemia. As Dr. Peters pointed out, sometimes this fear is yet another burden that patients may not want to experience. Still, despite this downside and a lack of change in patients’ glycemic control, Dr. Peters believes the study demonstrated the potential effectiveness of tailored educational tools. Ultimately, Dr. Peters called for a “DIY diabetes survival kit” to manage care for patients living in these communities. Among her proposed solutions were: (i) an accessible type 1 diabetes help line; (ii) simple, subsidized programs to provide devices to patients; (iii) unlimited monthly visits with diabetes educators or other health coaches from similar backgrounds; and (iv) diversity support by simplifying access and reducing language barriers.

• To illustrate the importance of having solutions come from patients rather than clinicians, Dr. Peters shared a particularly compelling story of one of her patients who works as a mechanic. He was performing fingersticks just once or twice a week. When Dr. Peters met with him, she realized that his hands were so brutally callused (from his day job) that it was extremely painful and difficult for him to take a fingerstick – the photograph was shocking. Her immediate instinct was to prescribe him a CGM; however, he shared that he has to “slither” on the ground between trucks during his job, which presented a sensor adhesion challenge. Walking between trucks instead was not an option, as the boss would fire him for taking too much time to get between the trucks. The ultimate solution came from the patient – he realized that he could wear the sensor under a lumbar support band that he already used at work, preventing the sensor from being ripped out. It was just one example demonstrating the level of troubleshooting that must occur to effectively integrate technology into patients’ lives.

• We were very distressed to hear some of the challenges associated with applying for diabetes devices. The system is simply not designed with underserved patients in mind – in just one example, Dr. Peters explained how order forms require an email address, which many of her patients do not have. She also explained that it can be incredibly burdensome to find device support and assistance in Spanish – one (not named) CGM company had a terrible Spanish help line experience, while another had a fantastic one. As she pointed out, Spanish is an extremely common language in the US – imagine what patients who speak less-common languages must undergo.

3. Dr. Roy Beck on State of CGM: Accuracy Improvements Approaching BGM; Product Comparison; US T1DX Penetration at 31% (2017), Projections for Strong Growth, Especially in T2D

Jaeb’s Dr. Roy Beck provided a very positive overview of the CGM category in the US, covering the four available systems, comparing their accuracy, providing the latest penetration data from the T1D Exchange (31% in 2017), and offering an optimistic look into the future. Many of his slides are enclosed below, and we appreciated seeing the system-by-system accuracy stacked up side-by-side. The main takeaway: all four CGMs are pretty comparable on overall % within 15%/15 mg/dl (87%-91%), with some differences in hypoglycemia (G6 leading the pack at 90%, FreeStyle Libre behind at 53%). Accuracy overall with BGM had multiple problems, of course, including things like dirty hands, that are no longer a problem. Dr. Beck showed some JP Morgan projections for CGM uptake in type 1 and type 2 diabetes over a five-year period, which projected US penetration in 2023 at 87% in type 1 and 27% in type 2 insulin users. Still, said Dr. Beck: “It’s hard for me to believe we’ll get to 87% penetration in type 1, but these 2018-2019 figures are not that different from what we see in the Exchange. I don’t know what they [JPM] base their projections on, but even if it’s an overestimate, we’re definitely going towards CGM as standard of care for type 1 diabetes – and that has implications for study control groups.” The Exchange numbers are higher than the average in the US, of course, as Dr. Beck has said many times. He reiterated the T1D Exchange CGM penetration data shared in DT&T in January, noting 31% penetration in type 1 as of 2017 – including 51% penetration in those <6 years old (up 10x from the prior data). We agree with Dr. Beck that 87% CGM penetration feels very high, even though the stepwise gains in penetration will continue or even accelerate in the next five years – given where all four companies are going and with the high potential for AID (pumps and MDI) to become the “killer app” for getting on CGM. We also believe that once AID is more mainstream, this will become something there is pressure for every type 1 at least to have access to. For the type 2 JPM estimate – that’s a major overestimate in our view, even if they are referencing only insulin-using type 2s. That said, we also believe any type 2 on prandial insulin has the same needs as those with type 1 and should have the chance to use the best technology.

• Providing a sneak peek at three CGM trials to report at this ADA – WISDM, CITY, SENCE – Dr. Beck highlighted high rates of Dexcom sensor use (90%-100%) at six months, mirroring other recent trials of CGM. The percentage using CGM at six months for 6+ days per week (i.e., nearly all the time) was 84% in WISDM, 59% in CITY, and 92% in SENCE. This use rates bode well for positive outcomes in these age groups, which tested CGM in 60-83 year olds (WISDM), 14-25 year olds (CITY), and 2-8 year olds (SENCE); we’ll see data on Sunday and Monday. For comparison, CGM use was 89%-96% over six months in other recent landmark CGM studies (DIAMOND – type 1, type 2, GOLD, HypoDE, IMPACT, and REPLACE). “Use is dramatically better than what we saw in the JDRF CGM trial ten years ago.”

• Looking ahead, Dr. Beck highlighted the expected 2019 launch of Dexcom’s G6 Pro (fully disposable transmitter) and the late 2020 launch of Dexcom/Verily’s G7; the under FDA review FreeStyle Libre 2 iCGM and in-development FreeStyle Libre 3; Medtronic’s plans to pursue iCGM designation and improve Sugar.IQ; and Senseonics pursuit of pediatric labeling for Eversense, iCGM designation, 365-day wear, and a “scanning” version without a transmitter. Dr. Beck also noted the LifeScan/Sanvita (Nova Biomedical) announcement from a few weeks ago, which aims to launch a CGM in North America and the EU as early as mid-2020. CGM use, he concluded, is becoming more widespread, and the future will see sensors that are smaller, disposable, and cheaper – facilitating more widespread use, particularly in type 2 diabetes. “The benefit is far exceeding the burden now.”

• “CGM is really becoming competitive with BGMs in terms of accuracy.” Dr. Beck highlighted a 2017 MGH comparative study of 17 BGMs, which found that MARD ranged from 5.6% to 20.8% (!), with nine BGMs actually reporting MARDs above  10%!

4. Dr. Steve Edelman’s App Talk Underscores How Little Time HCPs Have to Keep Up with and Talk with Patients About Apps; BUT, He Believes Augmenting Adherence is Biggest Opportunity for Apps

UCSD’s Dr. Steve Edelman gave an “unexpected” view of the “physician perspective” on diabetes apps, which can best be summarized in four words: We don’t have time. He illustrated this point in a series of slides that laid out exactly all of the things an HCP must accomplish in a diabetes visit (usually 20 minutes or less):

• Very quick hello with little or no time to schmooze, which is important to initiate or maintain a good HCP-patient relationship.

• Take a history and do a physical (H and P) for the maximum level of billing (required and monitored by administrators).

• H and P includes an HPI (history of present illness), PMG (past medical history), review of medications (dose and time taken), past surgeries, allergies, dietary information including amounts, times, nutrient information, etc., and exercise including frequency, duration, and intensity.

• “10 point” ROS (review of systems) which includes asking questions about any issues and or complaints relating to the skin, eyes, ears, nose, throat, heart, lungs, stomach, and musculoskeletal system.

• PE (physical exam) which includes a general assessment, looking into the patient’s eyes, ears, nose, throat, feeling the thyroid, listening to the lungs, heart and abdomen, feeling the pulses, conducting a neurologic exam and examining the skin. If a new patient, then a rectal is typically required.

• Document any new laboratory results and compare them over time, i.e., A1c, lipids, blood pressure, kidney function, etc.

• Assessment and Plan (A and P) which includes reviewing all the important issues about each problem and documenting the plan for the patient. Each plan needs to be discussed and agreed-upon by the patient (shared decision making).

• For PCPs, the A and P needs to be done for all health problems (diabetes, high blood pressure, cholesterol problems, asthma, atrial fibrillation, constipation, arthritis, etc.) PCPs have on average 15 minutes per patient.

• After all that, answer any questions, confirm that the patient understands the instructions, refill any prescriptions needed and other important paperwork.

• Most HCPs are under administrative pressure to see as many patients as possible and get the highest level of billing. "To get those level 4 billing codes, you have to document the things on my list!"

So, Dr. Edelman wonders, is there time for HCPs to also address diabetes apps? Based on the suggestions outlined in the draft joint ADA/EASD position statement on apps, Dr. Edelman suggested that there isn’t really time for HCPs to also address diabetes. The statement urges providers to: (i) Be knowledgeable about apps and their strengths/weaknesses; (ii) Encourage and support the use of digital health apps as appropriate; (iii) Provide training and regular updates according to standards set by national and international guidelines; and (iv) utilize patients’ health data to improve quality of care and health outcomes. Dr. Edelman hinted that these recommendations are a tad bright-eyed: “In the real world, we have no time to learn about apps, to suggest them to our patients, or to make sure that we’re suggesting the right ones that have passed regulatory hurdles.”

• All that said, Dr. Edelman posited that improving adherence in type 2 diabetes is “where the money is.” As in previous lectures, he called poor adherence the key factor driving the efficacy gap between real-world and RCT, and noted that apps can be very good “motivation extenders” once a HCP-patient relationship has been established. “Can [apps] improve empathy on the side of the healthcare professional? Can they improve trust on the side of the patient? Can they help patients put diabetes higher on their priority list? Can they improve health care outcomes and quality of life?”

• In Dr. Edelman’s ideal world, there is: (i) A regulatory body that can rapidly evaluate the published data on apps, clinical effectiveness, and safety – this is more or less the goal of the PreCert program; (ii) There are resources to help HCPs summarize the various approved apps for patients – AADE’s DANA diabetes technology web resource is aiming to help here; and (iii) HCPs have adequate time to discuss apps with individual patients – sadly, we don’t see much positive momentum on this one. (The shift to MDs/coaches/educators providing continuous remote care may be the only hope here, propelled by companies like Livongo, One Drop, Omada, Onduo, Virta, etc.)

• Both Dr. Edelman and Glasgow’s Dr. John Petrie said that, at the moment, they learn about most apps from their patients. This isn’t inherently bad, though it may set up a paradigm in which the only apps that come to HCPs attention are those embraced by the early adopter community, leaving those that can benefit broader populations under-utilized. 

5. FDA’s Dr. Lias Endorses Pushing Device Functions to the Phone, Including AID Controllers and Remote Bolusing; Comments on FDA’s Safety Communication on DIY Products

FDA’s Dr. Courtney Lias kicked off a session on diabetes-related apps, giving her typically strong endorsement of pushing functionality to users’ mobile devices. She clearly stated that FDA’s focus is on improving options for patients through better device integration (enhanced by mobile apps) and on incentivizing interoperability and open systems through new FDA pathways. Encapsulated by these focuses are the new iCGM and ACE pump categories, as well as AID controllers, which FDA is encouraging manufacturers to integrate into (or at least be compatible with) mobile devices. Regarding the prospect of a lower-risk iController  class, Dr. Lias’ slide said “We are evaluating whether special controls may exist to make standalone AID controller software class II.” Notably, at ATTD, she commented that “we’ll see, but it’s likely” when asked about the same prospect. (Dr. Alain Silk talked along similar lines at D-Data today, though his tone implied this is a matter of when – not if.) She also twice mentioned remote bolusing from smartphones, extolling the benefits and positing that, due in part to advances in cybersecurity, we should be able to see devices that do this within the next few years. We’ll be interested to see who comes to market with this approved first – Medtronic? Tandem? Insulet? Tidepool Loop?

• Dr. Lias alluded to the recent FDA safety communication that warns HCPs and patients against the use of unauthorized diabetes management devices. She explained that there was a recent adverse event related to device interoperability (i.e., a DIY system), and while the patient survived, no one is sure what went wrong to induce the over-delivery of insulin, and there was no company to take responsibility for it. This is exactly the type of scenario in which clear guidelines about regulated, interoperable products will behoove patients, doctors, and regulators. As we noted at the time, an important point is that diabetes is a very dangerous disease to live with; all AID devices carry risk (DIY or commercial), and the risks of open-loop, non-automated therapy are very high. At the same time, FDA is clearly focused on getting approved devices to the market faster that will hopefully fill many of the DIY community’s needs (e.g., Tidepool Loop) while ensuring their safety via regulatory oversight and the benefits of having an approved product (e.g., a responsible party to launch investigations into device malfunctions, cease sales of the product until the issue is resolved, and issue a recall if needed).

Selected Questions and Answers

Dr. Gregory Forlenza (Barbara Davis Center): A few weeks ago, comedian John Oliver did a bit on the 510(k) process, railing on the difference between FDA-cleared and -approved. I’m curious if there is a movement to change some of this.

Dr. Lias: In recent news, there has been some people that believe the 510(k) pathway should be updated to stay with the times. Under the current framework, you can compare your device with a predicate that’s been on the market since 1976. Also, there’s a perception by the critics of the 510(k) process that no data is needed, which is not true across the board and also not true for diabetes devices. The 510(k) process works well for diabetes technology. For the diabetes devices we work with, the data standard is the same for 510(k) and a high-risk device. That said, regardless of your point of view ofn the 510(k) process, our Center has some new programs to boost device safety, including around the use of old predicates (which actually isn’t an issue too much in diabetes).

Q: Clinicians need more time with patients. But we’re missing apps that talk to EMRs. We have the power at FDA to, when we approve pumps/sensors, to make it a point that they should talk to EMRs. That could prevent mistakes so I don’t have to manually put in basal rates, ICRs from pumps; I think that will save time and be safer, and ultimately benefit the patient.

Dr. Lias: I think everyone can agree it’ll be better, and it’ll help reduce medical errors less. I think once they have an EMR, that company could take in device and app data. Efforts are underway to standardize the EMR so they can have same pathways in to the workflow. However, we do already have devices whose software does integrate with certain EMRs in certain health systems. I think that’s a great goal we’re all working toward.

Dr. Mark Evans (University of Cambridge): I have a question about the OpenAPS safety alert. There are increasing numbers of people using these. For those of us working in or outside the US, what happens if I come across someone on one of these systems, and they have an adverse event, and it’s not something to do with the device? What are my responsibilities?

Dr. Lias: That’s a challenge. There’s no one to take responsibility. With medical device manufacturers, you’d know who to talk to. We’re trying to work toward a place where regulatory pathways are reasonable so all efforts are covered with responsibility, under a quality system, so there are clear lines of responsibility where adverse events happen, and clear lines of innovation. After the safety alert, the DIY community put out a statement saying they support tracking of that type of information – you could also contact that community and let them know. We also want to hear that kind of information. The type of event we saw, we’ve seen in early clinical trials for AID systems, but those were being evaluated in an investigation phase, and the software issue was caught before something dangerous happened and before the device was on the market. We’ve never seen this type of adverse event in on-market AID devices.

6. Real-World 670G Data in Pediatrics (n=92) and Adults (n=34) Show Increased Time-in-Range but Decreased Time in Auto Mode and Sensor Wear-Time; 30% Discontinuation Rate in Pediatrics at Six Months

Barbara Davis Center’s Ms. Cari Berget presented intriguing real-world, observational data on use of Medtronic’s MiniMed 670G in pediatric (n=92) and adult (n=34) patients. After six-months of use, pediatric patients (average age 15 years-old) saw time-in-range increase from 51% to 57% (+1.4 hours). A1c declined by an average of 0.8% (baseline A1c: 8.8%); however, there was no significant change in A1c for patients with baseline A1c <9%. Despite these overall improvements, time in Auto Mode decreased from 65% to 51% and sensor wear dropped from 80% to just 68%. Ms. Berget did not offer up a potential explanation for the apparent contradiction – it is possible that patients were naïve to pump therapy and would have reaped these benefits from initiating any pump. A similar story was observed in the adult population at three months: time-in-range increased from 67% to 71% (up nearly one hour) while time in Auto Mode decreased from 79% to 72%. A1c decreased by 0.5% (no baseline given) but there was no significant change for those with baseline A1c <7%. Relative to time-in-range in the 670G pivotal trials (64%-74%, depending on age group) and real-world data (71%; n=119,724), the younger BDC users seem to be having a more challenging time with the system, adults are similar.

• The pediatric group demonstrated a “pretty high” discontinuation rate of 30% at six months. As Ms. Berget pointed out, this rate is close to the 38% discontinuation rate that Dr. Gregory Goodwin presented at ENDO from a separate real-world observational study (n=93) of the 670G in children and adolescents. At the time, Dr. Goodwin noted that this rate stands in significant contrast to clinical trials finding withdrawal rates closer to 5% – there could be several factors at play here, including demographics, degree of follow-up/support, and patient expectations. Either way, Ms. Berget rightly called for future hybrid closed loop systems to focus on reducing system burden to increase sustainability. We expect factory-calibrated systems from Dexcom and Abbott – along with lower alarm burden systems – will be big winners in the pediatric population.

• Ms. Berget highlighted her team’s new clinical training program for implementing hybrid closed loop therapy in a pediatric clinic. The protocol, published earlier this year in the Journal of Diabetes Science and Technology, leverages a quality improvement process called the Plan-Do-Study-Act model and consists of: (i) A 2-3-hour group class for Manual Mode training, emphasizing CGM education and diabetes self-care techniques; (ii) A 1-1.5-hour remote class to initiate Auto Mode after one week of Manual Mode; and (iii) Three short phone calls during the first month, targeting re-education and insulin adjustments based on the patient’s CareLink upload. Ms. Berget shared that the 72 patients who participated in the program achieved 72% time in Auto Mode and 83% sensor wear-time. She noted that follow-up during the first few weeks of 670G use were vital to patient success – phone calls lasted on average just 16 minutes, and the most common topics discussed were pre-meal bolusing, correction boluses, and responding to system alerts (patients experienced ~6 Auto Mode exits/week, mostly related to hypoglycemia). Insulin adjustments were also common, with insulin:carb ratios strengthened by 24% in three-fourths of participants.

7. IDC’s Dr. Anders Carlson on The Need for Uniform CGM Metrics: Simplifying Therapy Comparisons for Patients, Translating Clinical Research for Providers, Streamlining Regulatory/Reimbursement for the Health System

IDC’s Dr. Anders Carlson provided a three-pronged rationale for the importance of uniform CGM metrics, claiming that they: (i) simplify comparison between therapeutic options for patients; (ii) serve to define ranges and translate clinical research using a common language for clinicians; and (iii) streamline regulatory and payment methods for the health system. No argument there! As Dr. Carlson pointed out, patients are already familiar with the concept of time-in-range and have ranked it above or equal to A1c in a list of factors that impact their diabetes. Moreover, in the context of the recent emphasis on shared decision making, it can be difficult to create a management plan based on a static measurement like A1c. Conversely, CGM metrics like time-in-range can be understood by both patients and clinicians. Lastly, Dr. Carlson highlighted the increasing shift towards value-based care using CGM-derived outcomes, referencing Medtronic’s recent outcomes-based arrangement with Blue Cross Blue Shield of Minnesota, as well as Onduo’s value-based contracts.

• Dr. Carlson highlighted the two consensus statements published in 2017 defining CGM thresholds for reporting hypoglycemia (<70, <54 mg/dl), time-in-range (70-180 mg/dl), and hyperglycemia (>180, >250 mg/dl). He also noted that one of the roadblocks preventing more widespread incorporation of these metrics is the need to translate time-in-range into something meaningful for clinicians. To this end, he shared analyses showing that a 70% time-in-range roughly corresponds to an A1c of 7%. For each 10% change in time-in-range, there is an ~o.5% change in A1c. He also pointed to the recent ATTD 2019 meeting on time-in-range, which focused on establishing benchmarks for CGM-derived metrics – the official recommendations will be presented in 2-LB this weekend.

• While Dr. Carlson acknowledged that there is still work to be done in gaining wider recognition of CGM outcomes, he noted many abstracts at ADA include “time-in-range” in the body – we counted 58 in a search of the online planner with either hyphens include or not. He believes more studies on time-in-range “will trickle, maybe flood in, over time.” Dr. Carlson pointed to Dr. Roy Beck et al.’s DCCT analysis linking time-in-range to microvascular complications and Dr. Robert Vigersky et al.’s study associating time-in-range with retinopathy as two key papers laying the foundation for this body of evidence. We’d also include the recently published article from Drs. Roy Beck, Rich Bergenstal et al. showing fingerstick-measured biochemical hypoglycemia in the DCCT was strongly associated with subsequent risk of severe hypoglycemia.

Diabetes Therapy Highlights

1. GOULD Registry (n=1,735) Paints Dismal Picture of Guideline Adherence for ASCVD Management in US: Only 6% on “Optimal Medical Therapy” for Lowering CV Risk; Just 15% with Diabetes on SGLT-2 or GLP-1; McGuire Suggests Cost as Main Barrier, Kosiborod Says Clinical Inertia

Dr. Mikhail Kosiborod presented baseline results from the Getting to an imprOved Understanding of Low-Density lipoprotein cholesterol and dyslipidemia management (GOULD) registry (n=1,735 with clinical ASCVD and (i) LDL ≥70 mg/dL or (ii) on a PCSK9 inhibitor), which found that a “soberingly poor” 6% of participants were on guideline-recommended optimal medical therapy. Optimal therapy was defined as a high-intensity statin, an antiplatelet drug or anticoagulant, an ACE inhibitor or ARB, and among patients with type 2, either an SGLT-2 inhibitor or GLP-1 agonist. Results were simultaneously published in Circulation. By subgroup, only 5% of those taking a PCSK9 (n=151) were optimally treated, likely due to high-intensity statin intolerance according to Dr. Kosiborod. This compares to 7% among those with an LDL between 70-99 mg/dL and 5% of those with an LDL ≥100 mg/dL. Unfortunately, he noted that this is likely a “best-case scenario” considering that practices participating in the registry likely provide a better quality-of-care than those that don’t participate. Also, while prescribed medications were documented in the medical record, there was no ability to measure patient adherence to those medications. This said, it does appear that the vast majority of participants were taking more than one agent, with 88% taking a statin (the other 12% were likely intolerant, according to Dr. Kosiborod), 87% taking an anti-platelet, 72% taking an ACE/ARB, and 71% on a beta blocker. See graphical representations of these data below.

• Only 15% of type 2 patients were treated with cardioprotective SGLT-2 inhibitors or GLP-1 agonists – and only 55% with metformin. Dishearteningly, Dr. Kosiborod labeled this as “a typical representation of this day and age in the US.” There’s no doubt in our minds that so many more could benefit from these agents, though we understand barriers exist in clinical inertia and cost. Indeed, session chair and renowned clinical trialist Dr. Darren McGuire (SAVOR-TIMI, TECOS, DECLARE) posited cost as the main barrier to meeting guidelines, while Dr. Kosiborod believes clinical inertia is the primary problem. He reminded the audience that only ~10% of the Registry received ezetimibe despite it being generic (see above). Also, fewer than half of all patients, including in all subgroups, were receiving a high-intensity statin, despite this being their guideline-recommended therapy. Seeing as only 55% of those with diabetes were prescribed metformin, it would certainly appear that clinical inertia is prevalent in the registry, though we think it likely that cost contributed at least in part to the low uptake of cardioprotective medicines. As a reminder, overcoming therapeutic inertia in diabetes is a hot topic of late: ADA hosted a symposium on the topic in November 2018.

• During Q&A, one audience member noted that these numbers should be treated as a baseline for new guidelines, considering that the registry ran from 2016-2018 and SGLT-2s/GLP-1s were only promoted within ADA/EASD’s treatment algorithm in 2018 (though were strongly incorporated into ADA’s 2017 Standards of Care). GOULD will run for another two years, until 2020, which should give a better idea of the effect of these guidelines. This should also give more time for expanded indications, such as lowered eGFR thresholds for SGLT-2 inhibitors, to gain more traction, though Dr. Kosiborod noted that this effect was negligible in this analysis.

2. Dr. Jean-François Yale Proposes a Simple Yet Novel Solution to Reduce Patient Costs: Cut the Higher Dose Pills in Half

In a comprehensive talk reviewing the clinical implementation of SGLT2 inhibitors and GLP-1R agonists, Dr. Jean-François-Yale urged clinicians to leverage price equivalence of higher doses into lower patient costs. After a thorough rundown of clinical trial data that has illuminated the impressive effects SGLT2i and GLP-1RAs on A1c, weight, and cardio and renal protection, Dr. Jean-François Yale turned to what he views as the last hurdle for these therapy classes: cost. Very few people with diabetes are actually on either of these classes (see this 2017 Diabetes Care article), a fact he attributed to often-prohibitive costs that makes their broad scale use as a first- or second-line therapy challenging. Dr. Jean-François Yale highlighted this barrier by contrasting the patient cost of a “classical” treatment approach that does not involve newer drug classes (i.e. using metformin, a sulfonylurea, and a DPP-4) against costs with an “innovative” approach using newer drug classes (i.e. metformin, an SGLT2i, and a GLP-1RA). The total daily cost of the classical approach would be ~$16.60/day compared to ~$43.60/day for the innovative approach (and we note that even the classical approach may be cost-prohibitive for many!); to our understanding, this doesn’t account for reimbursement. How might clinicians work around this barrier? Dr. Jean-François Yale detailed a simple yet novel strategy that he uses in practice that can make a dramatic difference for patients: Costs for low and high doses of SGLT2i pills are currently identical, and this price equivalence can be uniquely leveraged to help patients reduce the costs of SGLT2 inhibitors, especially if they only require a lower dose. Dr. Jean-François Yale recommended that clinicians prescribe the higher dose of the pill, and then have patients cut the pills in half and take them separately – effectively halving the price of their medication (see slide below) while still getting the full lower dose. Of course, we can imagine several barriers that might limit the effectiveness of this strategy: Will patients view pill-splitting as too much of a burden? Could there be potential for over-dosing if a patient forgets to split? An alternative is to use an SGLT2i-Metformin combination pill of the higher dose but only 1 pill a day effectively giving a low-dose SGLT2i at a low cost with half of the daily metformin. We’re intrigued by the approach and curious about how widespread its adoption could be – surely, for the right patient it could work! Even more encouragingly, Dr. Jean-François Yale pointed out that a similar approach can also be taken with semaglutide, for which the high dose (1.0 mg/week) pens are priced the same as low dose (0.5 mg/week) pens. Adding both of these “hacks” to the equation, the cost of the innovative treatment approach can be lowered to $14.90/day – actually lower than the $16.60/day for the classical approach!

3. Full Phase 2 Results for Zafgen’s New MetAP2 Inhibitor Show Progressive A1c and Weight Reductions with Higher 1.8 mg Dose and No Safety Signals

Zafgen presented full results from its 12-week phase 2 trial of MetAP2 inhibitor ZGN-1061, demonstrating progressive A1c reductions with 0.9 and 1.8 mg doses, progressive weight reductions with the 1.8 mg dose, and no adverse safety signals. These full results build on topline results released in January 2019. After 12 weeks, the 1.8 mg dose of ZGN-1061 gave a 1.1% placebo-adjusted treatment difference on A1c (p<0.001), compared to 0.6% with the 0.9 mg dose (p<0.001). Baseline A1c in these groups was ~8.5%. Importantly and encouragingly, these reductions were clearly progressive (especially for the 1.8 mg dose), strongly hinting at the potential for even greater A1c reductions over longer periods of time. A similar phenomenon was seen with weight, as the highest dose gave a 2.3 kg (~5.1 lbs) difference vs. placebo at 12 weeks (p<0.001), with no evidence of this effect waning (see graphs below).  We’re curious as to whether Zafgen could further explore an even higher dose of ZGN-1061 for even more pronounced benefits, especially given the 1.8 mg dose was well-tolerated (more on this below). Interestingly, Zafgen tested doses up to 4.8 mg in its phase 1 trial for the candidate, with no adverse safety signals observed. Questions of dosing aside, these are impressive results for the candidate, and we’re reminded of enthusiasm shared by Dr. Jim Gavin at TCOYD/The diaTribe Foundation’s 12^th Annual Diabetes Forum last year during ADA 2018: “This is a fascinating molecule originally explored as an anti-angiogenic drug for cancer. It’s been repurposed and is absolutely wonderful in terms of mechanism of action. It can actually reduce weight and glycemia in type 2 diabetes, and it also reduces hepatic liver scores in NAFLD, so this is a wonderful technology that combines well with other things and could be a whole new class.”

• In terms of safety, no treatment-related serious adverse events occurred and there was no elevation in D-dimer concentrations (a biomarker of thrombosis). Furthermore, there were no clinically meaningful changes in markers of coagulation and no events of venous thromboembolism. As a reminder, ZGN-1061 is currently on clinical hold in the US because of FDA safety concerns related to thrombosis concerns with Zafgen’s previous MetAP2 inhibitor, beloranib, which was discontinued in 3Q16. On this front, however, Zafgen just provided an update on this clinical hold, stating that the agency has acknowledged positive in vitro assays for plasma coagulation and tissue factor expression. Next up will be an in vivo assay to confirm safety, which Zafgen and FDA are jointly developing.

4. Phase 3 Results for PPAR Pan-Agonist Chiglitazar Show Superiority to Placebo, Non-Inferiority to Sitagliptin on A1c; Improved Safety Profile Compared to TZDs and Potential in NASH?

Results from two separate phase 3 trials of non-TZD PPAR pan-agonist chiglitazar vs. placebo and sitagliptin demonstrated significant and sustained A1c reductions and a well-tolerated safety profile. In the CMAP trial comparing chiglitazar to placebo, both the 32 mg and 48 mg doses achieved superiority on A1c lowering at 24 weeks – the higher dose gave a very notable 1.5% A1c drop, compared to a 0.5% A1c reduction in the placebo group. In the CMAS trial, which used Merck’s DPP-4 inhibitor Januvia (sitagliptin) as an active comparator, chiglitazar achieved non-inferiority on A1c at 12 and 24 weeks (see tables below). Very notably, chiglitazar had a clean safety profile; adverse events were balanced between treatment, sitagliptin, and placebo groups in the two trials. This is a bit reassuring, and the field will likely keep a close eye on chiglitazar’s safety given that PPAR agonists – BMS’ muraglitazar and GSK’s rosiglitazone, in particular – sparked FDA’s 2008 CVOT mandate after potential CV risk was identified in observational studies. On the other hand, pioglitazone was shown to be cardioprotective in the IRIS trial. Chiglitazar is currently under development by Chinese company Chipscreen Biosciences, and both CMAP and CMAS were conducted entirely in China; we imagine US development is not a priority (or even on the table), as we’re very confident FDA would require a full CVOT for the candidate. In less positive news from the trial, chiglitazar was associated with slight weight gain (~1 kg), consistent with trends seen with TZD treatment.

CMAP Results: Chiglitazar vs. Placebo

CMAS Results: Chiglitazar vs. Januvia (sitagliptin)

• It’s currently unclear how chiglitazar might stack up against other innovative drug classes and within treatment algorithms shifting emphasis toward outcomes benefits. To be sure, the A1c reductions seen with chiglitazar in these trials are solid (~1% treatment difference with placebo), but with tried-and-true DPP-4s both taking hold in China and soon going generic, it’s unclear what further benefits chiglitazar could provide for patients. On the other hand, there’s always something to be said for patient choice. It remains to be seen what effect chiglitazar may have on heart and kidney related outcomes, and we are currently unsure whether Chipscreen will pursue longer outcomes trials.

• It’s possible that chiglitazar could find success as a NASH therapy. Both of these trials demonstrated significant reductions in liver enzymes (namely ALT, AST, and TBil) with chiglitazar, hinting at beneficial action on the liver. Chipscreen does plan to conduct additional studies and analyses in this arena, and we’re cautiously optimistic about what these studies may find. KOLs have generally remained bullish on the use of TZDs (namely pioglitazone – currently off-label use) in NASH, and we’re excited to see what effects a PPAR pan-agonist like chiglitazar may have.

• As background, chiglitazar is a configuration-restricted PPAR pan-agonist, carrying moderate activity on PPAR- γ (TZD target), PPAR-α (fibrate target), and PPAR-d. This compares to muraglitazar’s PPAR- γ/PPAR-α affinity and pioglitazone’s and rosiglitazone’s more selective PPAR- γ affinity.

5. With DECLARE Data, Economic Modeling Study Projects $2.8 Billion in Potential Healthcare Savings with Dapagliflozin in CKD

Using renal outcomes data from the DECLARE trial (we’re awaiting a deeper dive on Sunday!), an AZ economic modeling study suggests dramatic renal-related healthcare savings with the SGLT-2 inhibitor dapagliflozin (Farxiga) – to the tune of $2.8 billion (or $285 per patient) over four years. These numbers were derived by projecting the risk reduction for renal outcomes with dapagliflozin to all 10.8 million people in the US with diabetes and CKD. As presented at AHA 2018, Farxiga drove a 47% relative risk reduction (95% CI: 0.43-0.66) on a composite renal endpoint of  40% decrease in eGFR to <60, end-stage renal disease, or renal death. The presenter, Dr. Phil McEwan (University of Swansea, Wales, UK) elaborated that on standard therapy the cost of treating 1,000 people with diabetes and CKD for four years is $460,277 vs. just $174,833 with dapagliflozin (a 62% reduction). Specifically the model projects a 46% reduction in costs associated with worsening CKD and a 67% reduction in ESRD-related costs with dapagliflozin therapy. Dr. McEwan further elaborated that this avoidance of end-stage renal disease and related healthcare usage would prevent 15,614 new ESRD cases and an incredible 30,184 person-years’ worth of dialysis over the course of four years – an enormous win for quality of life. To contextualize, in addition to being one of the most prevalent complications of diabetes (affecting 36% of people with diabetes in the US), CKD is also one of the most costly, accounting for 29% of diabetes-associated medical expenditures. In total CKD is associated with an average of $10,165  per person in yearly healthcare costs; this ranges from ~$2,000/year for milder stage 1 and 2 CKD, but skyrockets to >$82,000/year for stage 5 CKD, which involves renal replacement therapy (either in the form of dialysis or kidney transplantation). The renal outcomes trial for Farxiga, Dapa-CKD, is expected to complete in November 2020.

• During Q&A, an audience member correctly pointed out that the cost of four years of dapagliflozin therapy would likely outweigh the $285 in healthcare savings across that time period. Dr. McEwan agreed, but contended that Farxiga is likely cost-effective when considering the healthcare savings associated with its positive effects on glycemic control, weight loss, and heart failure in addition to CKD.

6. Harvard Group Predicts CAROLINA (linagliptin vs. glimepiride) Results Based on RWD: 9% Non-Significant Decrease in MACE in linagliptin Arm, 2.4x Severe Hypoglycemia with glimepiride

Dr. Elisabetta Patorno – a key member of the FDA-funded Harvard team that has set out to try to replicate results from 30 RCTs using real-world data (RWD) – presented her group’s predictions for the CAROLINA CVOT. Based on real-world outcomes data from ~50,000 propensity-matched type 2s in Medicare fee-for-service and large US commercial insurance databases who had recently started on Lilly/BI’s DPP-4 inhibitor Tradjenta (n=24,131) or sulfonylurea glimepiride (n=24,131), the researchers anticipate non-inferiority on three-point MACE (non-fatal MI, non-fatal stroke, and CV death). Of course, Lilly/BI has already confirmed this result in a February announcement, but the group specifically projects a 9% non-significant decrease in MACE with linagliptin vs. glimepiride (HR=0.91; 95% CI: 0.79-1.05). Dr. Julio Rosenstock, the session’s chair and one of the CAROLINA results presenters (on Monday!), jested, “Before you continue, how much do you want to bet?” Even more granularly, 373 adverse CV events are expected in the linagliptin arm (27 per 1000 patient-years) vs. 458 in the glimepiride arm (30 per 1000 patient-years). As previously noted, CAROLINA would seem to confirm the CV safety of glimepiride, though we’ll wait to hear more on Monday before drawing any conclusions. Though Lilly/BI didn’t share any data related to severe hypoglycemia or incident ESRD (end stage renal disease) in its topline announcement, Dr. Patorno’s group also made estimations here: (i) In the RWD set, severe hypoglycemia was more than twice as common in the glimepiride arm (HR=2.38; 95% CI: 1.79-3.13) – no surprises here, within 160 severe hypos in the glimepiride arm and 60 severe hypos in the linagliptin arm; and (ii) There was no significant association between either therapy and incident ESRD (HR=1.08; 95% CI: 0.66-1.79). We’ll be sure to compare these conclusions with the data on Monday. This exercise was taken on as a more rigorous evaluation of the group’s ability to replicate RCT data. Replicating data, and proceeding to design a RWD analysis, when the outcome is known is one thing, but doing so when blinded to the outcome may be a completely different animal.

• This group has a history of accurately recapitulating RCT data with RWD. For example, EMPA-REG detected an elevated risk of DKA in the Lilly/BI Jardiance (empagliflozin) arm vs. the placebo arm (HR=1.9; 95% CI: 0.2-17.2). In a follow-up analysis, Dr. Patorno and colleagues compared DKA risk in >75,000 type 2s using SGLT-2 inhibitors or DPP-4 inhibitors in the real world, arriving at essentially identical results, but with greater confidence (HR=2.2; 95% CI: 1.4-3.6). This result also turned out to be consistent with the elevated DKA risk seen in the CANVAS (HR=2.3; 95% CI: 0.8-7.2) and DECLARE (HR=2.2; 95% CI: 1.1-4.3) RCTS. There have been other examples as well.

• Dr. Patorno explained that one of the ultimate goals of her group’s research is near-real-time analysis of safety and effectiveness of newly-marketed medications based on databases. Database studies, such as the much-hyped EMPRISE study, allow for sequential evaluations as RWD accumulates over time. Confidence and precision increase as the numbers grow.

• Another clear end-game is to have RWD studies more frequently replace long, costly RCTs. During Q&A, Dr. Patorno noted this as a future direction, but added that “we are still learning, we’re at the very beginning of the process. We’re trying to move toward potentially trying to recognize situations where we can avoid trials.” From what we’ve heard, RWD may have the most promise in replacing control arms (“historical controls”), enriching RCT data, expanding indications or populations, and meeting post-market requirements. Some of these are already very much in use, but there will be a continued push to expand their applicability.

7. Dr. Drucker Rants on Reproducibility Crisis in Research, Suggesting Reallocation of Grant Funding, Transferring Clinical Trial Standards to Basic Science, and Embedding Reproducibility into Hiring, Promotion, and Awards

Vocalizing many of the sentiments from his 2016 Cell Metabolism op-ed, University of Toronto’s Dr. Dan Drucker sounded off on the reproducibility “crisis” in scientific research, stating that more than 50% of published research is not reproducible. Our society, as he sees it, most highly values the spectacular, the amazing, and the transformative over solid, incremental science. This breakthrough mentality pervades all elements of academia and research, from training (or a lack thereof) in fundamental scientific processes, to hiring, journal publication, promotion and tenure, and media coverage. Academics are therefore incentivized to bias their work toward hypotheses and ignore differing data. While Dr. Drucker underscored that this bias is more often present in basic science work than clinical science, due to the strict reporting standards when research is done in humans, other biases exist across the board. For example, TrialNet Chair Dr. Carla Greenbaum pointed out during Q&A that authors often put a more positive spin on their work than is warranted. In fact, an unusual peer-review experiment published in the British Journal of Anaesthesia last week underscored this inherent bias, with original authors of a paper stating that the results for stronger doses of an anesthetic were “reassuring” on mortality, while a re-write of the exact same data by a different author found the evidence inconclusive. Altogether, Dr. Drucker underscored that, without a fundamental shift in research incentives away from garnering headlines and toward sound methods, the scientific community risks further loss of trust from the public and funders: “If I was a funder and heard that only 50% of the work I donated to was reliable, I’d go build another football stadium or arts center instead.” To this end, he gave several suggestions on how to bend the curve:

• Reallocate 10% of funds each grant cycle toward replication and reproducibility studies. According to Dr. Drucker, we spend hundreds of millions of dollars (on the low-end) chasing down high-profile, sloppy science that turns out to be wrong. Instead, let’s flip that on its head and prevent the bad science in the first place. Further, he highlighted the opportunity that funders have in imbuing the system with a greater emphasis on reproducibility. All they must do is demand a strong, accountable track record in exchange for funding, rather than focus on misguiding metrics like impact factor, which unfortunately lends itself to the “groundbreaking” work that does not stand the test of time.

• Hold basic science to the same reporting rigor as clinical trials. Just like all adverse events and outcomes are reported for every participant in an RCT, Dr. Drucker suggested requiring that same level of reporting for animal studies. Unfortunately, he noted that most of the basic scientists to which he’s floated this idea are “appalled”, since they are so used to trialing the whole gamut of doses and conditions before passing on those most optimal to human trials. While Dr. Drucker sees nothing wrong with tinkering to find the best conditions, he did question why other scenarios are not reported as well. Taking things one step further, Dr. Greenbaum suggested randomization in preclinical studies as well. While this undoubtedly adds expense, she argued that having a statistically sound foundation to draw on for human trial decisions outweighs the cost.

• Embed reproducibility into hiring, promotion, and award processes. As it stands, tenure decisions are often made on similar grounds to publications (“wow”, impact factor), rather than the reliability of their results. Moreover, even though the Banting Medal for Scientific Achievement (of which Dr. Drucker is a recipient) includes reproducibility as one of its requirements, he believes it is often overshadowed by the claimed grandeur of a discovery. In other words, we need to celebrate reproducibility for it to gain greater importance.

• Clinical trialist Professor Nick Freemantle discussed how RWD could conceivably augment RCTs to widen their external validity, provided results are similar and bias is reduced with proper propensity-score matching. However, this isn’t always the case. As examples, he highlighted the tangible (>3 standard error) discrepancy between EMPA-REG and CVD-REAL on all-cause  mortality (see below). While both studies suggest significant mortality benefit with SGLT-2 inhibitors (which we believe to be the main takeaway), the discrepancy in the effect size suggests that they are “measuring something fundamentally different.” On a more salient note, Prof. Freemantle noted that the three released studies comparing next-gen basals Toujeo and Tresiba also had differing results. While DELIVER D+ and BRIGHT found no significant difference between the two in hypoglycemia past the first 12 weeks, the CONFIRM real-world study (n=4,056) suggested that hypoglycemia and treatment discontinuation were both more common with Toujeo than Tresiba. We’d add that the Novo Nordisk-sponsored, phase 3b head-to-head trial (n=1,609) of the two agents, which was announced on the company’s 1Q19 earnings call, also found overall lower hypoglycemia risk with Tresiba. Altogether, these studies paint a murky picture of the comparative effectiveness of these insulins, further clouded by issues of trial sponsorship, according to Prof. Freemantle. We maintain that the biggest conclusion from these studies should be that next-gen basal insulins are major improvements over their predecessors, but we agree that this analysis does demonstrate hurdles in comparing RWD to RCT data.

  • If you’re interested in this space, we implore you to keep an eye on Harvard’s ongoing efforts to replicate 30 RCTs using RWE. Dr. Jessica Franklin’s FDA-funded project has identified 40 replicable RCTs, including six large CVOTs, to figure out what types of questions are best answered with RWD. Professor Freemantle also noted that both FDA and EMA are very interested in understanding what ways RWD can complement RCTs in regulatory decisions.

8. Positive Results for Lilly’s Nasal Glucagon in Japanese Patients

Dr. Munehide Matsuhisa presented data demonstrating non-inferiority of Lilly’s nasal glucagon compared to intramuscular glucagon (IMG) in type 1 and type 2 Japanese patients. The study enrolled 72 people (33 type 1s and 39 type 2s; baseline A1c 7.7% and 8.1%) in a randomized, single dose crossover study in Japan. 100% of those in the nasal glucagon and IMG groups achieved the primary outcome measure of (i) an increase in plasma glucose levels to >70 mg/dl or (ii) an overall increase of >20 mg/dl from the glucose nadir within 30 minutes of receiving glucagon. Notably, all participants on either treatment actually hit both of these measures, indicating highly robust efficacy results. As a reminder, Lilly has filed nasal glucagon with FDA and EMA as of 2Q18, although we’re unsure of its current regulatory status in Japan – this trial would have been required for submission there. A decision by FDA is now expected in 3Q19, following FDA’s extension of its review time by up to three months.

• In terms of time to treatment success, nasal glucagon was actually significantly slower than traditional glucagon, but Dr. Matsuhisa contended that this difference was not clinically relevant. The mean time to treatment success was 12 minutes for the nasal glucagon group compared to 11 minutes for the IMG group. We tend to agree that it’s unlikely that this one-minute difference would be clinically relevant in most situations. Time to treatment success was also similar between those with type 1 and those with type 2, supporting nasal glucagon’s use in both populations.

• Nasal glucagon was understandably associated with several nasal-related symptoms, including runny nose (7%), nasal congestion (11%), and watery eyes (21%). Otherwise, overall treatment-emergent adverse events were relatively well balanced between the nasal glucagon and IMG groups (17% vs. 13%, respectively). Notably, rates of nausea were halved in the nasal glucagon group compared to IMG (6% vs. 11%) but total event numbers were too low to draw conclusions (4 events vs. 8 events, respectively).

9. Dr. Des Schatz’s “Argument with [Himself]” Over Whether to Implement Population T1D Screening; All Signs Point to “Yes,” Except There’s Still No Safe & Effective Intervention

University of Florida’s Dr. Des Schatz delivered a masterful and deliberately provocative address on the cost-benefit of screening for type 1 diabetes risk. Download his slides here. Currently, the ADA recommends screening for type 1 in asymptomatic children only in research studies in relatives of those with type 1. Widespread screening is not recommended due to the lack of viable interventions. At the end of the day, Dr. Schatz would not change ADA’s recommendations, though that would change as soon as: (i) the cost-benefit equation changes; or (ii) an intervention is shown to be safe and effective in slowing the progression of type 1 diabetes. He arrived at this conclusion after a thorough analysis of screening technology, clinical benefit, potential social costs, economics, etc., all built around the four requirements for public health screening. These four criteria, whether type 1 screening meets the requirement, and his accompanying comments are outlined in the following bullets:

• Cost/benefit to individual and society. (Type 1 diabetes? YES.) There is a strong case to be made for population screening, starting with early diagnosis to decrease morbidity/mortality; Dr. Schatz cited data showing that the rate of diagnoses presenting with DKA are significantly attenuated in the TEDDY study vs. registries, and that those diagnosed in TEDDY have better beta cell function, lower A1c, and decreased daily insulin doses vs. community controls in the first year post-diagnosis. The other clear upside of screening initiatives is they help us to better understand the natural history of pre-type 1 diabetes, gain insight into the etiology and immunopathogenesis, and identify individuals for prevention trials. Though the “benefits” outweigh the costs in Dr. Schatz’s view, there are also some downsides to screening. For one, there is no evidence that catching type 1 diabetes early actually leads to lower cost and complication burden long-term. Second, false positives take a serious emotional toll on families, particularly young parents. And then there’s the financial cost of screening: in screening to reduce DKA, the breakeven point is $1.00 per HLA screening test and $0.03 per autoantibody test, which is lower than current technology can achieve. 

• Disease detected early enough to intervene. (Type 1 diabetes? YES.) Screening tools can identify, with high fidelity, individuals who are in the pre-symptomatic stages of type 1 diabetes, which enables prevention (if/when such tools become available), DKA prevention, and earlier treatment initiation.

• Effective method for identifying those eligible for intervention (sensitivity, specificity, positive predictive value, low false positives). (Type 1 diabetes? YES.) Population genetic screening at birth, with follow up for 5% (autoantibodies, etc.) can detect 60%-80% of individuals at risk of developing type 1 diabetes. Autoantibodies (e.g., ICA, GADA, IAA) and genetics (e.g., DR/DQ genotypes) are highly predictive. Once an individual has 3+ autoantibodies, there’s a 75%-80% likelihood he/she will go on to develop type 1 diabetes.

• A credible intervention must be available. (Type 1 diabetes? NO.) A safe and effective preventive measure is the clear missing link that would make population screening for type 1 a no-brainer. On this front, we’ll have our eye on Sunday’s highly-anticipated phase 2 results for anti-CD3 agent teplizumab in those with ≥2 islet autoantibodies, as well as the Helmsley-funded GPPAD study which includes the screening of 30,000 babies and high-dose oral insulin therapy for those at high risk.

Questions and Answers

Dr. Marian Rewers (Barbara Davis Center): In the ASK study, we’ve screened many people for autoantibodies, and we have early cost-effectiveness data that I’ll share with you. The cost to detect a case of future diabetes in the general population is <$5,000 using the techniques we’ve developed for ASK. It is enough for the screening to pay for itself if we are able to lower rates of DKA by 20%, from 46% at baseline to 35%. That’s not a very ambitious goal. But that wouldn’t pay for screening in itself. In addition, we have to be convinced that by diagnosing kids earlier, starting insulin earlier, we are preventing some complications down the line. There is evidence that kids diagnosis without DKA have better A1c by 1% than those diagnosed with DKA. I think we are checking off slowly the boxes on the list you presented, and there is exciting cost-effectiveness data even with the tools we have today.

Dr. Schatz. I feel very mixed about this and that’s part of the debate. I have a couple of comments. Related to cost – it’s really in the follow-up, there are a lot of false positives. Lots don’t progress, and there are negative consequences of non-progression. We also really haven’t thought about education. If the population out there were better educated about recognition of DKA, about the  symptoms of diabetes, school teachers, ER docs pediatricians … 40% of these kids in DKA were seen by a pediatrician or family practitioner a week before they presented in DKA! The answer is by education, and spending less money, you could be cost effective. From all the other perspectives of disease, prevention, I’d agree we need to be doing these things, and hopefully a couple more of these studies will be able to delay the disease, because that’ll shift the pendulum.

Dr. Rewers: Would you say we are failing to prevent DKA? Why can’t we do as well as our Canadian, Australian, or Scandinavian friends? Why does it have to be 46%? Poster 1360 shows that the rates in Colorado have increased further, now they’re >50%. If we can’t prevent DKA using community education, maybe we should start screening.

Dr. Schatz: But in other countries, they have lower rates, and it’s not because they’re doing more screening. They’re not.

Q: What about the cost-effectiveness in a multi-ethnic population?

Dr. Schatz: I don’t have  an answer, but that’s a very, very good point. The incidence of type 1 in different populations is very, very different. So, you would have to think about targeted screening. I don’t think we’re far enough along the corridor yet.

10. Does Vitamin D Supplementation Prevent Type 2 Diabetes? D2d Study Shows a Trend Toward Benefit but Misses Significance

In the D2d trial of vitamin D (vitD) for prediabetes (n=2,423), supplementation with 4,000 IU/day of vitamin D3 did not significantly decrease risk of progression to type 2 diabetes compared to placebo (HR=0.88, 95% CI: 0.75-1.04, p=0.12). There were no significant subgroup interactions. See the full publication in NEJM and the Kaplan-Meier curve below. Of note, an exploratory per-protocol analysis identified a stronger trend toward benefit (HR=0.84, 95% CI: 0.71-1.00) when censoring (i) stopping of study pills (n=135 in vitD arm), (ii) initiation of diabetes or weight-loss medication (n=29 in placebo arm), and (iii) taking out-of-study vitD above 1,000 IU/day (n=63 in placebo arm) – indicating protocol violations may have dampened the efficacy of vitD. Indeed, the Kaplan-Meier curve reflects a dynamic outlined by Dr. Erin LeBlanc (Kaiser Center for Health Research) in her presentation of baseline characteristics. More patients in the placebo group than the vitD group started taking out-of-study vitamin D above 1,000 IU/day during the trial, and that trend grew stronger toward the end of the trial, peaking at 581 members of the placebo group in month 48 (2.4% more participants compared to the vitD group; 95% CI: 0.9%-4.0%). At the same time, the nonsignificant trend toward benefit with vitD aligns with two other RCTs of vitD in prediabetes: Norway’s Tromso study (n=511) achieved a HR of 0.90 vs. placebo (95% CI: 0.69-1.18), and Japan’s DPVD study (n=1,256) achieved a HR of 0.87 vs. placebo (95% CI: 0.68-1.09). In his presentation of study results, Tuft’s Dr. Anastassios Pittas pointed to the consistency between these trials and D2d’s primary outcome as an indicator of the results’ reliability. From where we stand, consistent trends toward benefit in three RCTs should be viewed positively, but also balanced against an apparently low magnitude of benefit: The rate of new-onset diabetes in D2d was 9.4 events/100 person-years with vitD compared to 10.7 with placebo (293 vs. 323 total events). For comparison, DPPOS has demonstrated significant reductions in diabetes incidence out to 15 years (HR=0.82, 95% CI: 0.72-0.93) with metformin – a similarly low-cost and simple approach to diabetes prevention. Data on glycemic parameters, as continuous variables, is forthcoming.

• D2d randomized 2,423 people with prediabetes, defined by meeting two of three ADA diagnostic criteria: A1c 5.7%-6.4%, FPG 100-125 mg/dl, or 2-hour OGTT 140-199 mg/dl). All participants were overweight and at least 30 years old, and the study excluded those taking diabetes medications or who were planning to begin obesity treatment; all received currently-recommended care for prediabetes (including recommended lifestyle changes), as well as for vitamin D and calcium intake. At baseline, the study cohort had an average age of 60 years and BMI of 32 kg/m², was 67% white, and was 55% male. Mean FPG was 108 mg/dl, A1c 5.9%, and 2-hr PG 137 mg/dl. While both groups started with a serum 25(OH)D level of 28 ng/mL, the vitD group climbed to 52 ng/mL at year one and 54 ng/mL at year two, compared to stagnation at 28 and 29 ng/mL in the placebo group. Retention in both arms was 93%, and median follow-up was ~3 years.

  • Both A1c and FPG were checked every six months and an annual 2-hour OGTT was performed to monitor for progression to type 2 diabetes. A type 2 diabetes outcome required meeting diagnostic criteria on two of the three tests. The study was designed for 508 diabetes events, assuming a fairly strong 25% risk reduction with vitamin D for 90% power.

• This high degree of vitamin D supplementation proved safe: There were no significant differences in prespecified adverse events of interest, including high calcium, low GFR, and kidney stones. As Dr. Pittas described, there has been concern over toxicity with achieved vitamin D levels above 50 ng/mL, and 4,000 IU/day is considered the upper limit of safe supplementation. To our understanding, D2d serves as the best evidence to date on the long-term safety of high vitamin D supplementation.

• D2d was sponsored by the NIDDK, ADA, Office of Dietary Supplements, and National Diabetes Education Program and run through Tufts. As a reminder, the study was undertaken in light of (i) observational studies indicating an association between vitamin D levels and diabetes and (ii) an NIDDK-sponsored pilot study of 92 participants over 16 weeks linking vitD supplementation to improvements in Disposition Index (adjusted insulin secretion).

-- by Adam Brown, Ann Carracher, Abigail Dove, Martin Kurian,

Brian Levine, Peter Rentzepis, Maeve Serino, and Kelly Close