Greetings from Bethesda, where day #1 of the 17th annual Diabetes Technology Meeting did not disappoint! We got our first taste of the clinical performance of Bigfoot’s AID system (and of Chief Engineer Mr. Lane Desborough’s ability to predict the outcomes through impressive modeling and simulation); saw the simplified user interface on Insulet’s Lilly U500 OmniPod PDM; heard Dr. Bruce Buckingham on smart pens (2017 launch timing still expected for Companion Medical’s InPen; listened intently to FDA’s Dr. Courtney Lias enthusiasm for interoperability (public forum possibly coming) and her view of key trends in CGM (factory cal included); heard a former FDA official (Dr. Zan Fleming) proclaim that “the day of reckoning is coming” for CDER (in reference to acceptance of non-severe hypoglycemia); and appreciated an enthusiastic talk on the value in CGM in type 2 diabetes (more Libre Pro enthusiasm).
One thing’s for sure – attendees seem energized by the agenda, which has a different tone than it did only a few years ago when there was perceived by many to be more “noise” between various valuable presentations (the eye is in the beholder of course). Digital and automation and even reimbursement are at DTM now, while factory calibrated CGM and AID aren’t just “research” projects anymore. Of course, some of the same questions are still coming up (especially adoption barriers), but overall, this is a meeting characterized by some greater optimism – that technology can take us much further than where we’re at today and that adoption barriers can be addressed, at least to some degree, and that reimbursement is moving toward value-based discussions, which should help the field.
The “official” meeting starts Friday, with sessions on AI for decision support, augmented reality (can it finally break out of the exhibit hall and into the patient experience meaningfully?) and innovations in automated insulin delivery. Read our conference preview here to see what else will be going on in Bethesda on days #2-3! For now, read our highlights from Day #1 below.
- Executive Highlights
- Top 8 Highlights
- 1. Bigfoot AID Topline Feasibility Results (N=20) – 65% TIR (70-180), 0.9% <70 mg/dl; Lane Desborough’s Remarkably Accurate Simulation Predictions
- 2. First Look at Screenshots of Well-Designed U500 Lilly/Insulet OmniPod PDM
- 3. FDA’s Dr. Courtney Lias: Interoperability Will Drive Innovation (Working With HCT, JDRF) + 5 Trends in CGM, Including Factory Cal
- 4. Dr. Bruce Buckingham on Smart Pens; Companion Medical On Pace for 2017 Launch; TypeZero Pilot Data
- 5. Treatment of Type 2s “Can Be Revolutionized” with CGM; “AGP is the Holter Monitor for Dysglycemia” – Dr. Eugene Wright
- 6. Dr. Fleming: Day of Reckoning Coming for CDER (With Respect to Acceptance of Non-Severe Hypo)
- 7. Senseonics Poster Details High Real-World Usage, Strong Glycemic Data in 56 EU Patients Over 90 Days
- 8. Alt Insulin Delivery Update: Oramed t3-Month Study of Oral Insulin Soon; Dance Completes 2 PK/PD studies of INhaled Insulin
Top 8 Highlights
A popular Bigfoot poster gave attendees a first look at the insulin automation capabilities of the Bigfoot AID system, which has been quite stealth to date. The multi-site, single-arm safety and feasibility trial enrolled 10 children (7-18 year old) and 10 adults to wear the system for 48 hours – including meal and exercise challenges – in a clinical research setting. The feasibility study (at the time) used the Dexcom G5 CGM and a control algorithm built into the tubed pump (the disposable pump body from Asante, the durable controller has been customized by Bigfoot). Mean CGM glucose value through the study was 164 mg/dl, as participants spent 65% of time between 70-180 mg/dl, 0.9% of time <70 mg/dl, and 34% of time >180 mg/dl. These results are quite consistent with other AID studies, though of course, outcomes from this might look different in a longer trial. In presenting the poster, Chief Engineer Mr. Lane Desborough only referenced them once, noting that the company moved quickly to get an initial safety and feasibility trial done ASAP. Rather, the main attraction was to Mr. Desborough and co.’s modeling work – prior to the human study, over 100 million subject-days of simulation were performed to evaluate algorithm candidates, tune parameters, and predict performance across a range of conditions. (Yes, machine learning can run scores of clinical trials in seconds.) Amazingly, the extensive modeling and simulation work allowed the team to predict almost exactly what the outcomes of the human study would be, anticipating an identical 164 mg/dl mean CGM glucose value and ridiculously close 68% time between 70-180 mg/dl, 1% of time <70 mg/dl, and 31% of time >180 mg/dl. The CGM plots pasted immediately below – with the smooth, black S-curve representing the projected and the red clouds representing the actual outcomes – show how close the Bigfoot simulation projections were. The company has now characterized its algorithm extremely well in simulations, and all for $5,000 (Amazon Web Service cloud fees).
- Mr. Desborough indicated that running a simulated trial cost Bigfoot just pennies per patient-day, whereas a traditional in-clinic closed-loop trial runs tens of thousands of dollars per patient-day – that’s more than a million-fold difference! As an added benefit, the company avoided the need to put people through the discomfort of clinical trials while the algorithm development was still in progress.
- Noted Mr. Desborough in a follow-up conversation with us: “The main takeaway for me is that simulation allows you to do things which would be impossible or impractical or unsafe to do in human trials. Control engineers perform experiments to develop models to be used in control algorithms. Once implemented, those control algorithms transfer variation from a place where it hurts (in our case, blood glucose) to a place where it doesn't hurt as much (basal insulin adjustment). Unfortunately “experimentation” and “closed loop control” have diametrically opposed objectives. The former seeks to maximize variation in the controlled variable (i.e., blood glucose) through carefully designed perturbations of the manipulated (i.e., insulin) and disturbance (i.e., carbs, exercise, stress) variables while the latter seeks to minimize variation in the controlled variable (i.e., blood glucose). Experimentation for the purpose of developing closed loop algorithms is thus caught in a dilemma: maximize variation for learning or minimize variation for safety and efficacy? Performing prospective experiments and characterizing behavior in simulations rather than human subjects allows us to avoid this dilemma.”
- As a reminder, Bigfoot expects to begin its pivotal trial next year with a next-gen factory calibrated Abbott FreeStyle Libre (launch possible in 2020, pending approval). Some have expressed skepticism that Bigfoot can’t possibly enter a pivotal so soon, as it only one feasibility study under its belt. This simulation data speaks to what the company has under the hood (hundreds of millions of patient days), and Mr. Desborough makes a strong case for cost effectiveness and predictive value with this approach. While it cannot simulate every use case, we believe modeling should play a larger role in the AID algorithm’s toolkit. We expect to hear more on the feasibility trial data and modeling tomorrow; Jen Block is presenting. Presumably Bigfoot’s pivotal trial will enroll a similar number of participants as Medtronic’s 670G pivotal, though we could imagine more or less would be included (more to capture a wider portion of the market and age groups; less because of the modeling approach).
- The simulations were run on Bigfoot’s vClinic, which simulates broad sources of blood glucose variation – change across populations, change throughout the day, change with activities and events, and change over time. In one such simulation that comprehensively included a missed meal bolus and an exercise challenge, 100 virtual subjects were sampled from the population distribution, each characterized by a unique combination of basal rate, carb ratio, insulin sensitivity factor, and open loop blood glucose distribution. Each subject was then simulated with and without AID system configuration setting mismatches – that is, with 27 different combinations of basal rate, carb ratio, and insulin sensitivity factor. Simulations even took into account unmeasured disturbance inputs, and included models for common pitfalls: carb counting error and CGM drift, dropout, calibration, and miscalibration. Wow!
- Will simulations become more important in diabetes technology over time? As a reminder, Dexcom relied on simulation data at the G5 insulin dosing FDA Advisory Committee meeting last summer. At the time, many speakers (including FDA’s Dr. Courtney Lias) strongly supported use of the simulation models, even though the some of the panelists (or at least a couple) were confused by them initially. Though not the total package, Dr. Claudio Cobelli explained that the simulations are powerful because they allow for thousands of trials that directly compare SMBG and CGM in the same exact virtual user and with all else equal, and because they cheaply enable investigation in certain scenarios that would be dangerous and unethical to study in humans.
2. First Look at Screenshots of Well-Designed U500 Lilly/Insulet OmniPod PDM
Insulet Medical Director Dr. Trang Ly showed screenshots of the U500 Lilly Omnipod PDM for the first time publicly and explained design considerations for the type 2 population. She explained how, after iterative UX development and human factors research, the user interface has been revised for simplicity, with a condensed home screen and easier navigation. Specifically, text size and contrast have been adjusted for readability, and certain more-complex information features have been removed or de-emphasized. Key information, such as last bolus and last blood glucose are displayed on the home screen, and time/insulin left in the Omnipod appear on the PDM (both images below). We noticed the U500 home screen looked quite different than the U100 Dash PDM we saw at ADA, which is intentional – these are specific for the U500 pump, which will be a separate product to Dash, hence the deliberately different color scheme. The greater simplicity of this U500 version is excellent. Insulet and Lilly have also developed a U500-specific syringe that reflects the actual units of insulin, eliminating the need for mathematical conversion. Based on the company’s user research, these adjustments will greatly benefit the target type 2 population – many of those interviewed were not familiar with insulin on board (they understandably thought it was how much insulin was left in the pump reservoir), duration of insulin action, correction factor, and insulin:carb ratio. This was good for us to see to better understand what has to take place in terms of education. Most who were already on U500 insulin didn’t fully understand why or know how it was different from what they had previously used. Disturbingly, the surveys revealed that many patients had limited training on their pumps, knowledge of just the basic bolus function, and some wouldn’t even touch the pump out of fear that they would “do something wrong.” As Dr. Ly said, they were “really petrified” of some of the more complex features. It is amazing to hear Insulet unearthing these insights from people with diabetes, designing for them, and then talking candidly about them. Of course, not all type 2s fall into this camp, but these anecdotal findings underscore the echo chamber effect we are increasingly reminded of – the “average” person with diabetes is a lot different than the people sitting in the audiences of digital health/diabetes tech conferences. At the same time, Dr. Ly and team seem to have totally embraced the valuable design thinking, iterative approach to patient-centric design to ensure safe and effective use. Dr. Ly didn’t present any day results from the recently-completed phase 3 trial – data is currently being analyzed, and per Dr. Ly’s AADE remarks, will hopefully be presented at ADA 2018. The U500 PDM is expected to launch in 2019, per Insulet’s last update.
- As we’ve come to expect, Dr. Ly noted that OmniPod is the only approved insulin delivery device on the market not reimbursed by Medicare. She acknowledged that there would be many likely U500 PDM customers in that population, and assured that the company is working on it. The 3Q17 call today shared confidence from CEO Patrick Sullivan, but did not give any timing update. One of the sticking points is whether the OmniPod should fit into Part B or Part D; Insulet is comfortable with either.
- Lilly’s Ms. Jennal Johnson and Dr. Ly actually split the presentation time in their discussion of the impactful partnership. The collaboration was announced ~4.5 years ago (!) and has been pushed back a few times. We’re glad to see the finish line (that is, launch) coming into view, as concentrated insulins in a pump (on-label) is a true unmet need; Dr. Ly shared that 25% of overall U500 insulin use comes from pumps today!
FDA’s insightful Dr. Courtney Lias provided an excellent update on five key trends in CGM and shared continued enthusiasm for interoperability. On the latter, she noted that “interoperability can promote innovation,” get patients products faster, speed regulatory reviews, and enable more choice. FDA is working on “smart solutions” for seamless device interoperability (“low hanging fruit”), and discussions are ongoing with stakeholders – including JDRF and the Helmsley Charitable Trust! Dr. Lias said FDA will “hopefully be announcing some public opportunities for discussions on that front. We’re trying to remove barriers that don’t actually help.” This is outstanding news and we hope it drives to actionable device communication standards that companies follow – the potential here is very big, as noted in our coverage on JDRF’s new open protocol initiative for automated insulin delivery. Dr. Lias also gave a nice update on five trends in CGM, highlighting better accuracy, fingerstick replacement claims, no calibration, implantable, and better use of data/integration. Read more on these below, followed by some of our favorite slides.
- Dr. Lias specifically mentioned the past year’s approvals of Abbott’s factory calibrated FreeStyle Libre (September) and Dexcom G5’s non-adjunctive claim (December 2016) to replace fingersticks, noting FDA is “very comfortable they will be beneficial.” However, both companies will be doing post-approval studies to ensure the fingerstick replacement claims are safe. We knew about Dexcom’s study, but this is first we’ve heard Abbott also has to do one. (As a reminder, T1D Exchange’s REPLACE-BG study showed non-adjunctive use of G4 is safe.)
- Dr. Lias was quite positive on factory calibration – this is a “movement in all sensors…Calibrations can provide greater continued accuracy; however, they also are a big source of error. Many people don’t do calibrations properly. Factory calibrated sensors, as long as companies are creating stable sensors with good performance, have potential to vastly improve patients’ quality of life.” Of course, this comment is not surprising after the FDA approval of FreeStyle Libre – which was not a given! This comment bodes very well for Dexcom’s no-calibration plan for G6, which has potential to launch by the end of 2018, per yesterday’s 3Q17 update. It also bodes well for Bigfoot’s plans to use factory calibrated FreeStyle Libre in its automated insulin delivery system (pivotal in 2018, potential launch in 2020).
- Regarding “Big Data” and integration, Dr. Lias gave several valuable examples of possible CGM-driven software/apps/decision support. “You have all this CGM data; what do you do with it? How do you use mobile platforms to optimize life with diabetes, share across devices, and give data to HCPs?” Examples:
- “CGM-based dosing calculators” (we assume this means meal/correction bolus advice based on trend arrows, carbs, and perhaps historical trends).
- Clinical decision support (we assume using CGM data to help HCPs titrate therapy)
- Mobile apps for coaching/lifestyle (e.g., last time you ate this meal, you needed this much insulin; you should take X units this time)
- Mobile apps for integrated and personalized tracking
- CGMs for type 2 diabetes (“a real trend,” noted Dr. Lias)
- Automated Insulin Dosing Systems (presumably both pump and injection-based systems that use CGM data to direct dosing– either in real-time or a few times per day for injectors)
- Regarding the above, Dr. Lias believes “these are going to be coming along quickly, and in various formats.” She admitted these can be hard to assess on the algorithm front, since all the algorithms are different from each other and there is no history like we now have in traditional automated insulin delivery. Still, her comments signaled (to us) enthusiasm for the potential of CGM data + software algorithms to drive smarter and safer dosing.
- Despite improved CGM accuracy/reliability, Dr. Lias believes the field has not hit a plateau. To this end, the FDA is really working with companies on CGM manufacturing standards to improve the reliability and accuracy of products. “How is the company assuring that when they release product, it performs the way it should? That’s harder than it seems.” Manufacturing is indeed the name of the game in CGM these days, as it drives product performance and cost – two major drivers of uptake. Abbott and Dexcom have both invested significantly on this front, and Medtronic is now playing catchup.
- On implantable, Dr. Lias did not mention Senseonics’ Eversense by name (currently under FDA review). However, she said “commercially feasible [implantable] products are being manufactured. We will see whether or not they reach the market in the next several years. We will see what develops.” She noted that the on-body footprint of implantable devices is often not much different relative to current generations, an area of opportunity. As a reminder, Senseonics’ Eversense remains under FDA review, with a possible FDA advisory committee expected in 1Q18. The product currently under review is the first-gen transmitter, which is indeed quite bulky compared to the 55% thinner transmitter that has already launched in Europe.
Stanford’s Dr. Bruce Buckingham was very excited to discuss the long-awaited smart pen, commenting that he believes Companion Medical’s InPen, which we’ve confirmed is still on track for a 2017 launch (two months left!), is going to be a “game-changer.” Dr. Buckingham showed a number of new-to-us screenshots of the InPen app (below), including the bolus calculator, and the “reverse bolus calculator” (input: blood glucose; output: additional grams of carbs to consume). In his review of the landscape, Dr. Buckingham called attention to the Timesulin dose capture device (now a part of Bigfoot), and the Echo Pen from Novo Nordisk (piloting in Sweden). Instead of arguing the usual case that clinicians need data from connected pens because it’s a huge gap in clinical care, Dr. Buckingham showed it. He displayed a series of sensor and pump downloads showing post-prandial excursions. “You might think you need a more aggressive insulin:carb ratio,” he said, “but it’s the wrong thing to do.” He then pressed the clicker and the bolus size and timing data appeared on the bottom of the screen. In each example, it was clear that making the patient bolus more aggressively would’ve been a very bad idea, because it was a matter of timing; they were bolusing after they had already started eating. Having this information makes it much easier (read: possible!) to make therapeutic decisions. Wearing a different hat, Dr. Buckingham noted that connected pens also make the job of a researcher much easier, because it’s very difficult to do a study in diabetes if you don’t know when and how much insulin was given.
- Dr. Buckingham also briefly flashed pilot clinical trial data from the TypeZero open-loop decision support system (DSS) pilot clinical trial in pumps and MDIs on the screen. Based on the (confusing) figure below, the TypeZero system seems to have reduced hypoglycemia and brought a tighter range of glucose values in the study population. Impressively, TypeZero’s decision support offers smart bolus calculation, hypoglycemia prediction, exercise and sleep advice, and in silico therapy optimization, and eA1c. Dr. Buckingham added that his group is conducting an ongoing clinical trial, along with groups from UVA and Mt. Sinai, with TypeZero’s inControl MDI Advisor. This is the first open-loop decision support system connected to a sensor that he has worked with.
- He also briefly reviewed the Esysta smart pen that recently launched in Germany. In one Esysta clinical trial of 215 participants (mean age: 60 years; 81% T2D), after ~14 months of observation, A1c dropped 0.9% (baseline: 8.7%). In the type 1 participants, mean total daily dose also decreased. Dr. Buckingham noted that the patients in the Esysta study who had more interactions with their providers saw the biggest drops in A1c, underscoring the concept that technology is best leveraged as a complement to the care from a provider. Given the scarcity of trials in this area, as well as discouraging results in some early trials with older memory pens (e.g., Danne et al., 2012 with Lilly’s HumaPen Memoir), we hope to see this field evolve as rapidly as closed-loop has. Connected pens seems like an exciting frontier for payers and pharma, who will get far better data on how drugs are used in real life.
Duke’s Dr. Eugene Wright Jr. beautifully asserted that treatment of type 2s “can be revolutionized with CGM.” This data has already been tweeted and re-tweeted! His persuasive talk shared clinical experience using Abbott’s FreeStyle Libre Pro, though he never mentioned the device by name. (The AGP reports in his slides were obviously from Pro.) Dr. Wright argued that professional CGM provides excellent insights for HCPs and people with type 2, including those on orals – identifying hypoglycemia and 24-hour glycemic profiles, making glucose-lowering therapy adjustments, initiating treatment changes, and as a behavior modification/conversation tool. He showed multiple reports of patients with A1c values of ~7%, which masked significant highs and lows that were only identified with AGP: “A hemoglobin A1c of 7% is not a hemoglobin A1c of 7%.” Hear, hear! Dr. Wright concluded that “AGP is the Holter monitor for dysglycemia,” a great way of putting the vision. His clinical cases showed clear value of CGM in type 2, though he did not address a big question for the field: patient selection for real time vs. professional CGM. Adam posed this question in Q&A, to which Dr. Wright was honest – he has only had experience with FreeStyle Libre Pro so far. “Your point is a very good one. Should everybody be able to see all their data all the time? In patients who are on sulfonylureas or insulin, I think it’s a good idea that they get to see it. Those on other agents that may or may not cause low blood sugars, it may be less important. Periodic and episodic monitoring, along with A1c, can be useful, and probably more cost effective.” We are huge fans of 24/7 CGM use – the real-time learning and feedback is unparalleled – though wonder how the type 2 market will segment out. Will some patients get just as much value out of professional CGM at key junctures, paired with HCP advice? Perhaps this is a study worth running…
6. Dr. Fleming: Day of Reckoning Coming for CDER (With Respect to Acceptance of Non-Severe Hypo)
In reference to FDA’s (CDER’s) resistance to acceptance of non-severe hypoglycemia as a key regulatory endpoint, Kinexum’s Dr. Zan Fleming, exclaimed “we need to get over this hurdle, and I think the day of reckoning is coming for [CDER’s Division of Endocrine and Metabolic Products]. The Division has been out of step, and has not accepted technology that has evolved to the point of regulatory quality for supporting therapeutic development.” We just learned yesterday that the EMA has approved a label update for Tresiba to reflect reduced risk for severe hypoglycemia vs. Lantus based on the DEVOTE trial, and Dr. Fleming doesn’t see how FDA will be able to resist this time in granting its first hypoglycemia claim on a drug label. Of course, this doesn’t answer the question of non-severe hypo yet, but it is a definitive positive step, as was the involvement of CDER’s Director Dr. Janet Woodcock at July’s Glycemic Outcomes Beyond A1c.
- Dr. Fleming’s Kinexum colleague, Dr. Douglas Muchmore, preceded his talk with a brief history of hypoglycemia definitions before proposing that 70 mg/dl stand as an appropriate alert level and 54 mg/dl be designated clinically meaningful and relevant for hypoglycemia and for use as a cutoff in clinical trials and for regulatory considerations. These are the bins that were agreed upon in July and will be published in a slew of upcoming papers. Dr. Fleming went further, suggesting that there is no need or practical way to prospectively validate CGM endpoints by showing the relationship between consensus-based CGM endpoints and long-term clinical outcomes. “It’d be a pretty big project, and I don’t see how you do it.” It would be more feasible, he added, to choose hypoglycemia endpoints on the basis of abundant available data and stick to them. “I believe you could hang your hat on a simple glucose level, which is justified by a large body of evidence, as a start for a reasonable regulatory approach.” Dr. Muchmore chimed in, citing evidence that counter-regulatory hormones kick in at 70 mg/dl, and cognitive impairment can set in at around 54 mg/dl. (Dr. Stephanie Amiel presented very convincing data on 54 mg/dl at EASD.) In a study of diabetic coma, of the patients involved, all had a blood glucose lower than 50 mg/dl at the time of the coma, save for one. To Dr. Muchmore, this is all the evidence that is needed – a blood glucose of <50 mg/dl doesn’t guarantee a coma, but the data is telling and does not necessitate a clinical trial. We agree! UVA’s Dr. Marc Breton commented that occurrence and depth of moderate/mild hypoglycemia also increases in days leading up to a severe hypoglycemia event.
- Dr. Fleming suggested that we actually have to be arbitrary in selecting thresholds: “Look at how we assess probability. We picked p=0.05, but why not 0.06? Why did we settle on 5% weight loss as meaningful? That’s the magic line. You want to know why? We made it up, pulled it out of you-know-where. Folks, we have to just hang our hats on what are to some extent arbitrary numbers.” Another defense of choosing glucose cut points is that the correlation between mild hypoglycemia and severe hypoglycemia is robust, and the occurrence and depth of moderate/mild hypoglycemia increases in the days leading up to a severe event. The field has done one better than pure arbitrariness, coming up with cutoffs (54 and 70 mg/dl) that are physiologically defined, and therefore likely tied to outcomes.
- Ultimately, everyone seemed to think the consensus cutoffs are viable, and tired of having the same discussion again. But we do get the feeling that the tide is turning and hope CDER hears the message – either behind the scenes, in meetings with companies, or by reading the upcoming papers. We hope the field moves to constructing a pathway to accept CGM-based data and outcomes for diabetes drugs.
A Senseonics poster shared real-world usage and glycemic results for 56 patients implanted with the Eversense CGM in Europe in May and followed for the 90-day sensor life. The big takeaway from the poster is that the users wore their transmitter 87.5% of the time in May (20.6 hours/day), 86.7% of the time in June (20.4 hours/day), and 86.0% of the time in July (20.2 hours/day). In the 180-day EU pivotal, median wear time was 23.5 hours per day over the first three months, and median wear time was 23.4 hours per day in the 90-day US pivotal. We’re not all that surprised by the ~three-hour decrease in real-world wear time vs. clinical trials. Wear time didn’t decrease meaningfully as the ninety days progressed (just over a 20-minute drop-off from May to June) – we wonder if this would hold true if the observation period were to stretch out another few months, and how many of the 56 opted to get re-implanted. We acknowledge that this group of patients adopted implantable CGM early and is likely therefore a highly engaged bunch, but the data is a positive for Senseonics. The poster also shared data on glucose levels: There was unfortunately no pre-sensor control data shared, but mean glucose came out to ~155 mg/dl for the 90-day period (decreasing from 159 mg/dl in May to 152 mg/dl in July). Average time-in-range (70-180 mg/dl) was ~63% (61% in May up to 64% in July). Meanwhile, time in the hypoglycemia ranges decreased from 1.6% in May to 1.2% in July for values <54 mg/dl (mean: 1.4%), and from 3.9% in May to 2.9% in July for values <70 mg/dl (mean: 3.4%).
Oramed CSO Dr. Miriam Kidron shared that the company will soon test oral insulin candidate ORMD-0801 in a three-month study, which will “hopefully bring us to phase 3 and registration.” We also saw preclinical data on an oral insulin+GLP-1 agonist candidate (ORMD-0901+ORMD-0801). As we understood it, the standalone oral insulin piece (ORMD-0801) is still years away in the best case scenario (won’t be on the market until after 2020), and an IND for the standalone GLP-1 agonist will be submitted soon. FDA has also requested food effect trials for the insulin product, which Oramed plans on conducting this year (i.e. how does the drug interact with food quantity and content?). We last saw phase 2b results (n=180; 28 days) for the insulin candidate in August. The trial showed a significant difference in the unconventional primary endpoint of mean change in nighttime glucose from run in (1.66 mg/dl for the ORMD-0801-treated group vs. 13.70 mg/dl for placebo, p=0.0117), as measured by a CGM. (We cannot ever recall a trial using such an endpoint.) As for the insulin+GLP-1 agonist pill, a small study of combination treatment in swine showed a solid glucose-lowering synergism between the two (see below). This is our first time hearing of an all-oral insulin+GLP-1 agonist product, and though it is clearly a long way away, the idea seems compelling if bioavailability and cost and efficacy are there. (Big Ifs, especially given how long Oramed has been working on this.) During Q&A, Dr. Kidron emphasized the decreased hypoglycemia risk (due to portal delivery) of oral insulin, and claimed that her company has an advantage over Biocon – Biocon tweaks the insulin molecule, while Oramed simply packages and delivers it. That said, she would be very happy if Biocon succeeds, probably for reasons of class awareness.
- In other news, we learned that Dance Biopharm has now conducted two PK/PD trials of its inhaled (liquid) insulin 501. Results suggest the formulation is slower than Afrezza up front (though more rapid onset than subcutaneous analogs) and has a longer tail, potentially making it a better option for coverage of mixed meals. The timeline we heard at DTM 2015 has obviously been pushed back – the company was previously planning for a phase 2b study in 2016, four phase 3 studies in 2017 (two in type 2, two in type 1), submission to EMA and FDA in 2019, and approval in 2020. We are surprised to hear that the product is still in development, given the challenges Afrezza has faced and the direction insulin pricing is heading.
- Said Dr. Lutz Heinemann: “If you are an insulin company and you aren’t thinking about having an automated system, you are stupid.” Dr. Heinemann said that oral insulin companies are apparently working with closed-loop researchers (as an adjunct therapy) and have data. An AACE symposium did share that groups at Yale and UVA are testing adjunctive Afrezza in the automated insulin delivery setting.
- On inhaled insulin, Dr. Heinemann doesn’t think it’s a dead horse, but does it have a bright and straightforward future? He’s not sure. He chalked up the Sanofi/Mannkind debacle to a host of factors, including a restricted label and poor marketing and advertising, but noted that Mannkind is active on the market on its own now and that patient satisfaction with inhaled insulin is very high. As a backdrop to his comments, in 2Q17, Afrezza sales grew 25% sequentially to $1.5 million.
--by Adam Brown, Brian Levine, and Kelly Close