Alrighty, thanks everyone for joining us today for our first investor and analyst event. Sorry for the technical difficulties and brief delay, but we're up and running now. Yeah, appreciate everyone joining on an early Sunday morning. By way of quick introduction, I'm Blake Beber, Head of Investor Relations at Beta Bionics. We have a great agenda for you today. I'm joined here by Sean Saint, our CEO, Mike Mensinger, our Chief Product Officer, Mark Hopman, our Chief Commercial Officer, Dr. Steven Russell, our Chief Medical Officer, and Stephen Feider, our CFO. I'd like to remind everyone that some of the statements you'll hear today are forward-looking in nature and reflect management's expectations about future events, our product pipeline, development timelines, and operating plans.
These forward-looking statements speak only as of today's date, and we undertake no obligation to update them to reflect subsequent events or circumstances, except to the extent required by law. You can refer to this slide, our disclaimer slide, and our investor event deck, as well as our SEC filings for more information on forward-looking statements. Both the replay of this webcast, as well as the supplemental slide presentation, will be available on our website for one year following the conclusion of today's event. Information recorded during today's event speaks only as of today, June 22, 2025. If you're listening to the replay, any time-sensitive information may no longer be accurate. Without further ado, we can get started here. I'd like to introduce Sean Saint, our CEO.
Thank you, Blake. Hello, e verybody. I can assure you that it will be properly reflected on Blake's review. Anyway, thanks for coming today. I know it's early. We've got a bit of a jam-packed agenda. We'll try and move through it quickly. We're going to leave a ton of time for questions at the end. We're going to start today with a commercial update by Stephen and Mark. I'm going to talk a little bit about not our algorithms specifically, but algorithms generally, and the differences between the algorithms, primarily because I think there's a lot of misinformation out there. I want to make sure we're all kind of singing from the same sheet of music on how to think about the differences in these algorithms. We'll talk about our iLet real-world experience, of course, with Dr.
Russell, and then, of course, the patch pump demo, probably why maybe a few of you were here. We are leaving it at the end so you have to listen to the rest of us. We are going to leave a lot of time for Q&A. With that, I am going to turn this over to Stephen.
Awesome. Thank you. Are we good? All right. Hey, everybody. Yeah, as Sean said, I'm Stephen Feider, and I'm Beta Bionics CFO. There's just two quick things for me this morning. Number one is a housekeeping item. It is our policy to not comment on the current quarter. Nowhere in today's presentation or in any Q&A are we going to comment on Q2 performance. The second thing is we actually launched a really cool product here at ADA this weekend. This new product is called the Bionic Portal. What it is, is an online healthcare provider portal. If you're interested in seeing the healthcare provider portal or the Bionic Portal in action, stop by our booth sometime today, and someone from the Beta Bionics team will happily show it to you.
There is one feature that I want to highlight here about the Bionic Portal. When we're presenting our clinical data or the clinical outcomes on the iLet, we like to use a particular graph to do this. What this graph does is it takes, again, patient population in the real world, and we take that population and we look at their hemoglobin A1C right before start on the iLet. We then compare that, and we take that starting A1C and compare that to the glucose management indicator. It's called abbreviated GMI, or their approximate A1C, excuse me, after they're using the iLet. Again, we take the starting A1C right before your start on the iLet and then compare that to the glucose management indicator, the approximate A1C after using the iLet. That clinical comparison is the strongest and most used sales tool that we have.
Our field sales team uses it every single day when we're talking about the iLet and showing the clinical outcomes to healthcare providers. Now, what the Bionic Portal does, and as of this weekend, is it makes those clinical outcomes, like that particular data I just described, accessible to any healthcare provider, any clinic who's prescribing the iLet. It does this in an online platform that's easily accessible. It's updated in real time, and it's in a format that's very easy to understand. We believe this is the first of its kind product like this that exists in insulin pumping. Here's just a snapshot or a screenshot of what it looks like. This is a particular—what you're seeing here is a particular clinic's patient population that's on the iLet.
What we show, again, on the left side of the screen is their starting A1C kind of bucketed, and then on the right side, their resulting glucose management indicator or their approximate A1C. A very cool product. Just in conclusion, the so what of this, like why does this matter? Demonstrating great clinical outcomes is core to what iLet’s value proposition is. We believe that by making these results more accessible for HCPs, that it will build even more confidence in our product and lead to even more high-volume prescribers of the iLet. Again, stop by the booth if you are interested in seeing this thing in action. That is all I have, all for me. The next thing is Mark Hopman, our Chief Commercial Officer, is going to tell you about our pharmacy business model. Mark.
Excellent. Thanks, Stephen. Yeah, I have one slide to go over with you today. We're going to walk through our pharmacy strategy and how it impacts the key stakeholders that deal with that. In early 2014, I accepted a newly created role at Dexcom, which was tasked with spinning up their pharmacy business. At that point in time, Dexcom had had zero prescriptions ever filled, zero revenue ever generated. About eight years later, when I left, we had done well over $1 billion of sales through pharmacy, and as you know, that's a major channel for them today. There was no analog at that time to copy. It had never been done before, so there was a lot of trial and error. We had some good success.
What I'm doing here in leading is we're distilling that down to the best practices that we learned during that and replicating it here. As we just announced on our Q1 call, we raised guidance for pharmacy new patient starts to 22%-25% for the year, so we're very pleased with how this effort is going. If you look at the overall strategy, there are three parts to it. First is the PBM agreement. This is getting on the menu. Many companies will state how many patients they have under contract, which means they're on formulary at a particular PBM. While we definitely look at that metric, it's not the metric we care the most about. We care more about how many new patient starts there are.
In this process, you have to take a deal to the PBM that meets their strategic goals and sign the contract with them. That does put you on the menu, but it does not make your product be ordered off that menu. The next part of this is health plan adoption. This is where they actually go in and model and look at the financials. They look at other things as well, but the financials are the most important things. Again, in the Q1 call, we talked about how Prime Therapeutics had adopted the Ascent formulary. We have seen some nice pickup from that, but that is the concept of the health plan adopting the PBM coverage.
Equally important to this, though, is you have to have operational excellence to execute against that coverage, and it's very difficult to straddle a pharmacy benefit and a DME benefit at the same time. We are not going to give away our secret sauce on how we do that, but we are very confident and very proud of what we have set up internally to identify which benefit the patient lives in and then help shuttle them down the appropriate pathway. From a stakeholder perspective on these new patient starts, just quickly looking at the different important players, you have the user. For the user, you have streamlined insurance approval. They get on product much faster. It is also lower out-of-pocket cost for our particular product to a patient. Commercially speaking, they pretty much never pay for the hardware. It is free to them.
They pay $25 a month or $300 for a total year of our product. That includes if they're in a deductible. It includes if they have to get the pump that year. That is a huge change. For diabetes, a very expensive disease, and they're used to paying the first year of pump therapy several thousand dollars, and then each subsequent year, a thousand or more. For us, we can confidently say for the vast majority of our patients, it's $300 or less per year. That lower out-of-pocket cost is big. Lastly, there is no pump jail. On pharmacy, there is no pump jail.
Pump jail on the DME side is you're locked in for four years because of that large outlay that the insurance company has made on your behalf, and they need the time to recoup the clinical return on investment that they paid for that patient. Patients hate that. No U.S. consumer other than maybe a house and their car sign up for something for four years, especially in a piece of technology that's going to evolve quickly. From a provider's perspective, those things are important too, but getting the patient on product quickly and also less administrative burden. On the pharmacy side, prior authorizations in many instances are done electronically where they do a 180-day look back for rapid-acting insulin in the patient's profile, and then they'll make an immediate pay decision. Whereas on the DME side, pretty much everything is still done manually.
Also, the no four-year commitment is big to them. From a payer perspective, and I'm going to head we get a lot of questions about this, so we're going to take it head-on. Is you know the economics of the pharmacy side versus the DME side. Over a four-year time period, it is about three times more costly. Why would a plan, when they're modeling this, decide to put the benefit on the pharmacy side? If you look at their population of patients on the product, it's a population that decays with two variables. The first variable is attrition from the product. If you're paying this is their economics. This isn't the economics Stephen usually talks about for us. From their economics, they're paying between $4,000 and $5,000 for that pump on the DME side.
If a patient's on it for a day, a week, a month, a year, they have not had time to get the clinical benefit, return on investment for that $4,000-$5,000. They put in place these mechanisms of pump jail. Imagine trying to land a message if you're health plan A that says, "Oh, you got a pump two years ago, even though you weren't our patient, and so we're not going to pay for it now for you." That's a horrible message to land. They hate that. That's one decaying variable. The second decaying variable is churn, what they call churn in the industry. Here's a situation. Imagine you're UnitedHealthcare. You just paid for a pump, and then two weeks later, that particular patient changes jobs and they move to a different insurer.
Now you, as UnitedHealthcare, just paid $5,000 for something that Aetna is going to have that patient for the next couple of years. When they do the modeling and they look at the pay-as-you-go model, they take the fact that there is zero upfront cost with a higher monthly cost to them versus that huge payment upfront and then the lower monthly cost. That is the number one thing they look at. They also like the less administrative burden on the prior authorization, but most of it is financial. We just raised our guidance on that. We are finding that sweet spot of where those numbers cross over for them financially, and they are adopting. Lastly, for us, it is faster access to users.
It's also a bigger addressable market because we can now go after other people who are in pump jail because it does not matter on the pharmacy side. The pay-as-you-go economics for us, it is about 3x the revenue generation over a four-year period than the DME side. We will take more questions at the end if there are on that process, but with that, I will throw it over to Sean.
All right. Thanks, Mark. Thanks, Stephen. All right. As I said earlier, I'm going to talk a little bit, not so much about the specifics of our algorithm, but the differences between them, and hopefully introduce kind of a framework and some words that we can all use to better understand these algorithms and, frankly, that you all can use to ask me questions, to ask my friends at other companies questions, to really understand what it is we're talking about. Because there's a lot of obfuscation that goes on about these different algorithms, people trying to make them look the same, when sometimes they are, and in other cases, they're actually quite different. I want to start with this. A closed-loop technology really has two stakeholders: the user, certainly. That's obviously the one we build this for, but also the healthcare provider.
Those are different concepts, and they require different words and different portions of the algorithm, frankly. Let's start with users. Continuum. Highly engaged algorithms. This is traditional. This is where we started. And then no engagement required on the right. We already have words for this. Hybrid closed-loop systems on the left. These are hybrid. They're not fully closed-loop because they require your engagement. Makes sense. On the right, fully closed-loop requires absolutely no engagement from the user. Where's the line between those? We'll get to that. On the left side, hybrid closed-loop, things like carb counting definitely need to be done. Things like manual correction boluses, temporary basal rate adjustments, extended boluses, all those things have to be entered into the pump manually. As you move to the right, you might start to see some simplification.
You might start to see pre-programmed carb boluses, fixed mealtime dosing, those sorts of things. Go a little bit farther, you're going to get to meal announcements. This is where iLet is today. You all know that scheme. I won't repeat it. It doesn't yet constitute fully closed-loop because those meal boluses are still required. Our label asks you to do it. Our contention is that the definition of fully closed-loop, and I think that this is consistent with what I've heard others talk about, is that you do not need to meal announce. They are truly optional. Maybe the feature is still there. Nobody's saying that you have to take away the ability to meal announce in order for it to be fully closed-loop. You just have to have them be fully optional. If you don't want to do it, it's totally fine.
It's completely in agreement with the labeling and the instructions. Hopefully that makes sense. We're not quite there yet. We do believe that we're the farthest to the right along this continuum, and we're going to show data in a little bit about the product being used in that mode. In other words, people ignoring the label and not meal announcing, which is their right and happens on every system out there at some level. I said two stakeholders. The second one is a healthcare provider. This is definitely an orthogonal concept. Let's show it orthogonally. Now here, we also have two ends of the spectrum. We have static, and we have adaptive. Again, let's show what that means.
The job of the healthcare provider, of course, is to set up and program the pump and make sure it works best for that person at all times. We are going to start with calculating the initial settings and programming them into the pump, and then keeping that pump good for that patient every three months. As we head toward adaptive a little bit, you might start to see setup wizards. We are seeing those. That is good. That helps with the initial setup of the pump, but it does not remove the requirement to review the data and modify those pump settings every three months. Go a little bit farther. You might start to see things like adaptive basal rates. That is certainly a little bit of a work reduction. Again, every three months, you are still in those pump settings. You are still reviewing the data, and you are still modifying them.
Then you get to adaptive, where there are, in fact, no settings to be adjusted by the healthcare provider. Of course, iLet. Okay. Now with those two concepts, I want to talk a little bit more about this portion, which is, again, the setup and management of the pump, not the use of the pump, and how difficult that is. What a healthcare provider is typically going to do when they start you on a pump is they're going to take your weight. Let's pretend this is a statement of medical necessity here. They take your weight, your total daily dose if they have it. They're going to utilize the rules of thumb or the ACE pump guidelines or the ACE guidelines, excuse me, to calculate all of the initial settings that need to be programmed into that pump. They program it into the pump.
This is the step here that a setup wizard could help with. This is our schematic example pump. It has five settings on it: basal rate, insulin sensitivity factor, insulin to carb factor, insulin to carb ratio, glucose target, and duration of insulin action. In this particular case, we've shown it over four time settings. You can use many more. You can use fewer. This is reasonably typical. What do we have here? We have 20 settings that need to be set up in this particular pump. Again, an example. Decent amount of work. What I want to show and what I hope to convince you of is that these settings are important and that they're difficult. Let's talk about what each and every one of the settings does.
Not all 20 because I'm not going to go through all the time settings, but you get the point. To do that, we're going to look at the dosing of insulin, the actual math associated with dosing of insulin. It's actually pretty straightforward. Every five minutes, that pump is, by the way, this math is exemplary. It's not anybody's exact math, but you'll get the point. It is how it works. Every five minutes, you're going to look at however much insulin you're going to get from your basal rate. You're going to take your blood glucose, subtract target, because that's how far above target you are. Divide that by your insulin sensitivity factor. We call that correction insulin. You're going to take the carbs you eat and divide that by the carb ratio. That's your mealtime bolus.
You're going to subtract from that any insulin on board you may have. That's insulin still in your body, still being metabolized, hasn't done its job yet, hasn't brought your blood glucose down yet. We get that number from a fairly complex equation that comes from your duration of insulin action. Now you can see how each and every one of those settings impacts the actual insulin delivery that we see every five minutes. You add that all up, you get your dose. What does that look like in the real world? Trust me, I have a point here. We're going to take our example pump, and what we're going to do is highlight each one of the settings that's active in a particular five-minute period so you can see what's going on.
We're going to look at a day in the life. This particular user starts at 180. You can see on the left, four of the five time settings are being used every five minutes. On the bottom, you'll see that the insulin delivery—no, that's not going to work. Oops. Let me back that up. Sorry.
Is this going to work?
Yeah. Down here, we're going to show insulin delivery. The bottom of the chart, you can see that blood sugar's coming down. Insulin delivery's kind of stable. Everything's sort of fine. But you're using four of those time settings each and every five-minute period. We get to 6:00 A.M. Those time settings switch over to the 7:00 to the 12-hour time. In this case, the insulin delivery's gone down a little bit because blood sugar was coming down. Now we get to a meal.
The user had to manually enter this, of course. We get a big bolus. That fifth setting, the insulin to carb ratio, has come into play, and we continue on. Blood sugar is now going up. The correction controller is going to start kicking in more insulin automatically. We get to near noon here. We are going to have another meal, bring in that fifth setting again, and continue on. You get the point. In this case, the user's blood sugar did not come down all the way, and they asked for a correction bolus. Now another meal. You can see all the interactions of the product by the user, but you can also see what each and every one of those settings was doing all day long.
Here's, in this case, the user's blood sugar went down after that last meal, and they asked for a temporary basal rate to probably help eliminate any impending hypo. There's the first day. What's interesting about this is you can actually look at any particular portion of this day and say, "Wow, they kind of went high after lunch." It's entirely possible that the carb ratio's wrong around lunchtime. It's a relatively straightforward thing to do. Bear with us because we can't actually do that. We put that day in memory. We go on to the next day. This day, we start more like 90 instead of 180. Other than that, the general things that are happening, here's a correction bolus, a meal, another meal. The day looks similar and yet almost completely opposite. Get to the end of that day.
Now that day is in memory. The problem with this is that the healthcare provider is in charge of not looking at these two days, but looking at data more like this. This is 14 days of data. We lovingly call this a spaghetti chart for obvious reasons. They need to look at this and deconvolute each one of those 20 settings for that user and try and modify them. In the best-case scenario, there's going to be some obvious trends in here. More commonly, it looks something like this, or it's very hard to do that. Hopefully, that's a good explanation of why trying to figure out these time settings every three months is an extraordinarily hard thing to do. That's what iLet is trying to do. That's what adaptive closed-loop algorithms are trying to do. Remove the settings completely.
Take away the necessity of staring at those spaghetti charts. We think they'll still look because everybody wants to know that their patients are doing well, but you do not have to is the point. That is what iLet is about. I guess the next important concept is we just showed 14 days of data because 90 days would be overwhelming. The healthcare provider is traditionally going to see you, as I have said several times, every 90 days. They certainly cannot see you every five minutes. They cannot see you every meal. iLet can. How fast can we adapt? This is a graph here of the predicted GMI that you get after starting. You can see down here at the bottom, after zero, one, two, seven days of use. What does it tell you? It tells you within a couple of days, it is adapted.
A couple of days. And then it looks completely static after that. But do not confuse that with the fact that the algorithm is static. The outcome is static. What that means is, as you change as a person, you change your diet, you start an exercise regimen, whatever it is, the system is staying with you automatically. You are not waiting for your next healthcare provider and a re-kickoff of another set of setting evolutions. The system does that automatically in real time. Okay. Hopefully, I have convinced you that these settings are hard, which means you need kind of a specialist to do that. Actually, a serious specialist. Call that an endocrinologist. I think we all know that, what, 40%-50% of the pump market is excuse me. 40%-50% of people with type 1 diabetes are managed by an endocrinologist.
These are the people with the skill sets to do that. There is a reason that primary care providers are not managing pumps very commonly. I have just shown it to you. The question, I guess, is, how easily can a primary care provider manage a patient on iLet, or how successfully can they manage a patient on iLet, or any fully adaptive system? Here is the answer. Endocrinology, we see an average reduction of 1.6% from baseline A1C down to follow-up GMI, the way Stephen described earlier. Primary care sees the exact same. Least surprising result ever, right? Because there is absolutely nothing to do. That is what a fully adaptive algorithm gets you: access to the other 50%-60% of people with type 1 diabetes in this country. Okay. Let us bring back those first two concepts that I talked about earlier.
There's the hybrid versus fully closed-loop continuum, and there's the static versus adaptive continuum. Obviously, you see we get kind of four quadrants. Our competitors, by and large, today are static hybrid algorithms. These are great systems. Nothing wrong with it. It's just what they are. We are trending toward fully closed-loop. Not there yet, but an adaptive system. Obviously, we've heard that they are intending to move toward fully closed-loop. That's great. We agree. We are doing the exact same thing. What I hope you leave here today with is an understanding that once you say fully closed-loop, that's not fully closed-loop in all regards. There's static fully closed-loop. There's adaptive fully closed-loop. Unless we get to fully adaptive closed-loop systems, we haven't opened up primary care. We haven't saved the physician any time, and etc., etc.
With that, and this new framework that I've laid out, I'm going to turn this over to Dr. Russell to talk about our first two years of experience on iLet and give us just a little bit more history on how this has been going. Dr. Russell?
Thank you, Sean. I'm going to tell you a little bit about what those algorithms, those adaptive algorithms, are able to accomplish in the real world. First of all, let's start with the background, the situation as we found it when iLet was launched. This is a histogram from data published in a paper looking at an insurance database. This was a large database. They looked at everybody who worked with that insurer, was covered by that insurer, and looked at their A1C at a point in time.
What you can see is that there is a large spread of baseline A1Cs as a reference point. The American Diabetes Association and all other professional organizations recommend that you try to achieve an A1C below 7% to minimize the risk of complications. For a long time, we've reported that in the population at large, only about 20% of people are able to achieve that goal. That means that the rest are at much higher risk of complications. You can see in this particular population, it's about 23% have an A1C below 7%. There's a long tail with people with much, much higher A1Cs all the way out to greater than 14%. How does the population that are started on the iLet compare to that? This, in black, is a histogram of the people starting on the iLet across the first two years of use.
You can see it's somewhat similar to that overall population. It's a little bit worse. In other words, the population has shifted to the right. People have worse glucose control before they start the iLet than in this insurance database. I like to point out that that's probably because those were people who had commercial insurance. We know that people with Medicare and Medicaid tend to have worse glucose control than people in that commercially insured population. If you look, you can see that we have more people in the 9-10%, 10-11%, 11-12%, and so on, population bins as compared to that insurance database.
I think that's because the iLet has, as I'll show you, we've earned a reputation for being able to manage glucose even in people who have traditionally had a very hard time maintaining good glucose control because of the amount of automation that we have. This is that rainbow graph that was described earlier that Stephen talked about earlier. What this shows is dividing people's baseline A1Cs into bins and then looking at where they wind up on the iLet. You can see there's a broad range of baseline A1Cs all the way from 4%-6%, people who have really almost unphysiologically low A1Cs, all the way up to 14%-19%. Actually, our highest baseline A1C for someone starting on the iLet is 19.1% at this point. I have to admit, I didn't even know it went up that high.
We have seen that in our baseline data. You can see that the overall average baseline A1C is 8.9%, a little bit higher than the average baseline A1C in the population at large. What do we achieve with the iLet? You can see that everybody comes down into a much tighter range. We go from an overall baseline A1C of 8.9% to an average GMI of 7.3%. Not only do people with very high A1Cs get large reductions, everybody winds up in a very tight distribution around that new mean. That is because the iLet is adaptive. If your glucose is too high on one day, the iLet will automatically increase the basal dosing, the correction dosing. If you are getting too high of a postprandial excursion, the iLet will automatically increase the amount of meal insulin it gives you on the next day.
The iLet is continually adapting to your changing insulin needs. If you're having too many lows, the iLet will automatically back off on insulin dosing to try and mitigate that. Everybody winds up in this very tight distribution around the mean automatically without any intervention from a healthcare provider. That adaptation process is literally happening from minute to minute. Every five minutes, the iLet is adapting. That's something that, as an endocrinologist, I simply can't provide. I see patients every three months, every six months. I do calls with them in between times if things are going off the rails. There's no way I can provide daily input. If we go back to this distribution and ask, how do we change the game for these folks?
If I just go to that population schematic that I showed you before, and then I superimpose on it people on the iLet, you can see how that population changes. Before, there were about 20% of people who had A1Cs above 10%. On the iLet, it's zero. Zero. And those people were at very high risk for complications of diabetes. You can see the whole distribution shifts down dramatically. Most of the people are now in that 7-8 bin. Second most common group is in the 6%-7% A1C bin or GMI bin. And we have very few people above 8%. It's a total of 6% or above 8% GMI on the iLet. This is a remarkable sort of population health level effect. My goal in doing this, I've been working on this project a long time.
I've been involved with developing these algorithms since 2006. The goal was always to make an impact at the population level. Previous technologies that require so much input from the healthcare provider and where the outcomes are dependent on the skill of the user and how much attention they can devote to their diabetes management, they primarily were helping people who were already doing pretty well. If you had the skills to already do pretty well, you could get on these hybrid closed-loop systems, and you could do even better. We weren't really touching that big tail of the population that may not have had great numeracy skills, may not have had great executive function, may not have been able to focus and spend their time, or may just not have wanted to.
They did the math, and they thought, "I want to spend more time thinking about my life than I want to spend thinking about my diabetes, and I'm willing to take a risk, a long-term risk, to do that." With the iLet, they do not have to. Now, there is also subgroup analysis that we have done. One of the questions you might ask is, what about people coming from MDI? This is the first time they are using a pump. If we look at people coming from MDI, they have a worse baseline A1C than our population at large, 9.2% average A1C. They come down to a GMI of 7.3% on the iLet. You do not have to have expertise. You do not have to be a pumper to get these benefits from the iLet. This is a rainbow diagram to show that the general overall picture is the same.
People with very high A1Cs get very large reductions in A1C to GMI. People with better glucose control at baseline see less improvements, and they come down to a slightly better ending GMI. Overall, everybody winds up in that very tight range because of the automation of the iLet. What about people coming from other AID systems? Although the majority of the people coming onto the iLet are coming from MDI, we do have quite a significant number of them who are coming from hybrid closed-loop systems, other competitive AID systems. You can see that they do better at baseline. They have an average A1C of 8.1%, quite a bit lower than the population as a whole.
Those AID systems, those hybrid closed-loop systems are providing a benefit, but they still get significant reductions of A1C to GMI when they go onto the iLet because the iLet is relieving them of the necessity to be so involved in their diabetes care. It is removing that dependency on their skill, their judgment, how often their care provider can engage with them. Again, you see this rainbow diagram that shows you the pattern is very similar. We are still getting benefits in those people coming from hybrid closed-loop systems. This data was mentioned previously. Endocrinology practices get great results with the iLet. Remarkably, primary care practices also get fantastic results with the iLet. You can see that the number of people coming to us from primary care practices is relatively small.
It's almost 14,000 people in this database from primary care practices, only a little over 1,000—oh, sorry, from endocrine practices, and only a little over 1,000 from primary care. That is a tremendous opportunity. Our sales staff has primarily been focusing on endocrinology practices, as you might expect. We've been trying to learn more, going to meetings, trying to learn more about the primary care space. That is clearly a big opportunity for us. More than half of people with type 1 diabetes in the United States are managed by primary care doctors, and they're primarily managed with multiple daily injections because the hybrid closed-loop pumps are just too complicated. These are the rainbow diagrams that correspond to those bar charts. You see the same pattern over and over again. Everybody coming in within a fairly tight range. It's remarkable how similar those are.
They're almost identical coming from endocrinology and coming from primary care practices. What about fully closed-loop? We teased a little bit that we were going to talk about fully closed-loop. We have looked at our population. We now have quite a large number of people on iLet, and they have a range of behaviors, as you might expect. Some of them, they do the meal announcements as we suggest that they do. They announce all their significant carb-containing meals and snacks. It's simple. We think that's why a lot of them are able to do it. They just say, "It's usual for me, breakfast usual for me, or breakfast more than usual." Many of them choose not to do so. We have about 16% of iLet users who are announcing very infrequently, on average, once every three days.
This is over the entire time period of this database, this two-year period. On average, we've got about seven months of follow-up data for people in that cohort. We're talking about one meal announcement every three days over seven months. This isn't cherry-picking the one day that they didn't announce a meal and then aggregating that data. This is over the entire period. Those folks are getting remarkably good results. They have a worse baseline A1C, 9.4%, than the population at large. They don't do quite as well on GMI. They come down to 7.5% instead of 7.3%. That is a remarkably small difference when you're considering that they're almost never using the meal announcement feature. They're almost in fully closed-loop right now. In addition, I didn't mention this before, but our hypoglycemia rates are very low with our overall population.
That is also true of people using this system in a nearly fully closed-loop mode. 83% of them have both time less than 70 below 4% and time less than 54 below 1%. They actually have very similar hypoglycemia metrics to our population as a whole. Taking it a step further, we looked and we found that we have 530 users who have not announced a single meal in the last 21 days. We looked at what is their outcomes over that 21-day period. They went from a baseline A1C of 9.4 to a GMI of 7.4, full 2% reduction of A1C to GMI, even though they are not using the meal announcement feature at all. An even larger percentage of them are meeting all those hypoglycemia goals.
You may hear data from another AID company where they say, "Well, we looked at the days when people did not announce meals, and we have these amazing results." They do not tell you what percentage of days those are. Those are just isolated days, plucking a day out of a whole period of time where the person did not announce a meal. I mean, the first thing I think is, "Well, maybe that was a day where they fasted, or maybe that was a day when they did not have too many carbs." It does not tell you anything. We are reporting continuous days, whole patterns of behavior. That top group, these are for an average of seven months that people are using this way. I think it is much more believable that we are actually seeing good performance for people using it in that manner.
This is that rainbow diagram again for that group of people. Now I am going to pass it over to Sean and Mike to talk about a new initiative that we are working on that we think is really exciting.
Thanks, Steven. I am only going to take a second here and just introduce Mike. Mike is our Chief Product Officer here. What Chief Product Officer means is, yes, Chief Technical Officer. It also means defining what our products are going to be and ensuring they get all through the process. End-to-end products here at Beta Bionics. I do not know anybody better to do this. Certainly the world's expert on our patch pump program, which Mike is going to introduce right now. Mike.
Okay. Good morning, everybody. Great to be with you today. Before we actually get in and show anything on the patch pump program we're working on, we're going to take a second and talk a little bit about our design philosophy here at Beta Bionics. We know that diabetes, I think most of you know, is such a huge burden. It's one of the highest burden diseases that exist. People, you can't escape this burden. It follows you to work. It comes home with you. It's there while you sleep, even follows you on vacation. Our mission here is to lower that burden. And iLet has been in a class on its own to be able to remove manual correction calculations, manual carb counting, all these tasks that you need to do in diabetes.
Every minute that you spend on diabetes could have been spent with your loved ones, with your hobbies, any way you choose to spend it. Diabetes is the last thing you want to be doing. That is how we design products. That is how the algorithm has been designed, and that is how we approach every problem we do. When you look at other products, you may see a list of marketing features, square wave boluses and dual wave boluses, all these different things. We are defined by less, a lack of features, because we want less time, less things to do, and less tasks. When you see what we are going to present, that is how we came up with the designs that we come up with. With that, the best testimonial that we have had so far was an eight-year-old girl on iLet.
She said, "Today was the first day that I forgot that I had diabetes." Super powerful. That is what we strive to achieve, and that is what motivates everyone inside of Beta. With introducing our patch pump in development, Mint by Beta Bionics. Here is our pharmacy starter kit. Mint will come through the pharmacy channel available. Mark talked about the advantages through that channel there. The starter kit will come with, we will get into the architecture, their user guide, and everything they need for their first month of supply. We are going to do an example change here for you in just a second. Before we get into that, I wanted to get a little bit about the architecture and explain the product.
Remembering that user burden framework, we analyzed multiple different potential architectures, and we chose what was the best balance of lowering user burden and also minimizing environmental waste and other factors. It is a two-part design. We have a reusable controller, which we call the Mint controller. And we have the part that is disposable that the user throws away every three days, which we call the Mint cartridge. It is a 200-unit patch pump. It is intended to be used for three days. At three days, you may have changed your patch pump in the middle of the night, and now it is not a convenient time to change it at the end of three days. We are going to include a grace period so you can choose to change your patch pump within the next 12 hours when it is convenient for you.
You don't have to wake up in the middle of the night and change it on its schedule. The electronics include all of the electronics, the motor, the Bluetooth radio, and that great adaptive iLet closed-loop algorithm already built in. Even when you're not near your phone, the system is in closed loop and manually calculating all of the doses automatically for you. It will be controlled by an iOS or Android smartphone. We're going to update the iLet app so you don't have to carry an extra device with you, another controller. It'll already work in the device you already have in your pocket. Importantly, it will never require charging, which is a huge burden in any potential architecture. On the disposable side, of course, we have the 200-unit insulin reservoir. We have the batteries in there that enables no charging.
We have a very comfortable 4.5 mm cannula that minimizes insertion discomfort. Some of the advantages of this architecture here, first and foremost, the adaptive closed-loop iLet algorithm. This on its own in any patch pump form factor, even if we completely screwed it up, would be a compelling product on its own. To have all these outcomes that Stephen showed in a patch form factor is going to be a tremendous market opportunity for us. We are very excited about that. Getting the same or better outcomes with a lot less work in a patch form factor, very excited to see. We did not want to stop there. When we look at the user burden around the hardware itself, the majority of it comes in the patch change process. This is something the user has to do every three days. It is a lot of steps.
It takes a lot of time. We looked at re-envisioning that process, and that is what we will show you today. The important tips here that you will notice are, number one, you do not need to interact with your phone at all post-initial setup. The change process, even if your phone is downstairs in the kitchen, you can do it all without the phone. We also have that very small 4.5 mm cannula to minimize the insertion discomfort. We have tried to remove every single step, which we will go through in some detail in a second. We did not stop there. As Mark mentioned, we are also thinking about the user experience of how to get the product and the user experience of the provider and lowering the burden to get the product from your provider through the pharmacy process.
It minimizes insurance approvals, minimizes paperwork, back and forth between the doctor's office, and also out-of-pocket costs, getting it down to, on average, $25 a month. Tremendous user experience improvement through the pharmacy. We are surrounding it with a robust digital ecosystem. Having compatibility with the latest sensors, including G7 and the Abbott Libre 3 Plus, having that controlled through the upgraded iLet app that is already on the phone in your pocket. The Bionic Circle is a remote monitoring platform that we launched last year that enables parents and caregivers to have a full view of your glucose, your insulin delivery if you have announced meals, what the battery level is, or if it is time to change your patch, and how much insulin you have remaining if we need to change your patch early if you run out of insulin.
Surrounding all of that with a robust digital ecosystem. Before we actually go do the demo, quick overview of what you're going to see so you have some context. Five simple steps here. Step one, peel the patch pump off. Step two, you take that Mint controller out of the old cartridge. Step three, you fill up the new cartridge with insulin. Step four, you attach the Mint controller to the new cartridge. Step five, you apply the pump and insert the cannula. Very, very simple process. What you're not going to see here is involvement from the phone, which saves a lot of steps. We'll talk about that. No Bluetooth pairing required. You do that once every two years, not every three days anymore. No charging required ever. We'll go through this process here. I'm wearing a Mint right here.
For the demo, we're just doing saline. We're not pumping insulin. Inside of the Mint pharmacy pack here, we have two five-packs of the refills. We're going to pull out one of the new Mint. I've prefilled my syringe here with saline just to save some time. First step is to pull off the old product. We're going to take out the controller. We're going to open the new cartridge from the sterile packaging. We're going to fill it with insulin, saline today. My eyesight has gotten bad with age. The user can fill this with the desired amount of insulin. Not everybody needs 200 units over three days, so you don't need to fill it completely full. We're going to take off the liner, apply the Mint controller. The Mint will be initializing.
You can hear the beeping. It's hard to hear it. This is automatically priming and getting the system ready for delivery. We apply the Mint to the arm, follow the adhesive, and insert the cannula. Oh, sorry. Trigger guard, and then insert the cannula. That's it. Very quick, very seamless, very easy.
That was pretty cool, Mike. I feel like you glanced over a few things, though, so I want to make sure we're very honest and open with everybody about all the steps that we're doing. You didn't disconnect the old session. So we're going to stop that.
Yeah. The beauty is because we have the Mint controller that follows you from patch to patch, you do not need to go and have that step to tell the old one you are done with it. If you forget to do that on other products, it will be screaming at you on the table or in the trash can if you forget to switch.
It's like a scream of death?
It's a scream of death, yeah. No more of that. We're done with that experience.
All right. Cool. Good point. You did not pull out your phone to pair the radio because you have to Bluetooth it and wait until that.
Yeah. Wouldn't it be annoying if every time you got in your car, you had to pair your phone to your car? That would be annoying, right? We do not have that anymore. You pair it once every two years, and then you're done. That's a step you do not have to do. You do not have your Bluetooth settings cluttered with hundreds of patches after a year.
Looking forward to that. I'm familiar with a different experience than that, to put it that way. Yeah, but with these two-part systems, traditionally, there's a big criticism that one's on the charger, and when you pull it off, you have to pull it out. I didn't see you do that. Or you have to take the one that you pulled out, put it on the charger, and wait for an hour. You kind of jumped over that.
Yeah. Charging is such a hassle. If you're traveling, you forget your chargers, you would forget your extra. Architecture doesn't require it. We did debate that possibility, but charging is too high of a burden, too easy to make mistakes. That is why the batteries are included in the disposables.
Good idea. So what I'm familiar with, you tend to have to verify that it's on your body and then insert the cannula. And then there's a whole process. Again, you kind of seem to skip over that.
Yeah. Again, with this architecture, that was another. So you take the Mint controller, you install it in the new disposable, it automatically knows it's a new disposable, and it makes itself ready for delivery automatically. You don't have to tell it to go.
I also did not see Mint's. Is that thing really real?
Yeah. It is the smallest cannula that will be available with the 4.5 mm. Very tiny, very comfortable.
Yeah. I'm pretty familiar with the bite you get ordinarily.
Yeah. No ticking time bomb in the insertion process anymore.
Yeah. You got to love that one. Mike, I'm pretty familiar with patch pumps generally. You build them, you put the software on them at the beginning of the process. They got to go all the way through the build process. They got to fill the channel in McKesson, Cardinal Health, whatever your distribution channel is, and then ship to the patient. Then you have three months of supplies. They're in your closet. To get from end to end, you're going to take six months. For me, the challenge there is that if there's an upgrade to, say, some new sensor that may come out, that could take six months for me to get that software or that feature set. How are you going to handle that?
Yeah. Other products have had long delays because of the distribution channels and the supplies you may have at home. You may have several months. If you can't do a remote field update, then you have to wait for that new feature or the new sensor support. Just like with iLet, when we added compatibility with Dexcom G7 as an example, we were able to roll that out through remote firmware update, and users could update the next day after release and immediately have that benefit. With Mint, that's exactly the plan. You'll be able to do the same thing as with iLet. It's just open your iLet app and do a firmware update, and you're up and running with whatever new feature or sensor compatibility that we just announced.
You just said you need your phone. Apparently you need your phone for something.
Yeah. You do need your phone for a few things.
Anything else?
Obviously, you need your phone for setup. It includes some beautiful setup wizards designed by our amazing product development team. If you want to announce your meal, you do that through your phone, and you can receive alerts and alarms on the phone as well.
That's the only time it's required?
Absolutely.
Carb counting required?
No carb counting on iLet ever.
Pretty cool. You want to show us?
We can show you how we announce a meal, yes.
All right.
Let's see where the cord go. Here we go. This will be our updated iLet app.
I'm jealous of that blood sugar.
Yes. Thank you. Working pancreas is my secret now.
Let's see. Can we see it on the feed?
It's plugged in. Always a technical hiccup.
I can see it. The app's working. I got to switch over from slides. Maybe wiggle it a little. Worked before. I think it's make sure you didn't change the source. Okay. While we're doing that, I'll show an overview of what the process looks like here. As soon as we can get that going, what the user does is they're going to open up the new iLet app that'll be updated to have the new features available for the patch pump. Announcing meals through the phone, receiving alerts and alarms through the phone. They would tap the meal announcement button. The user simply selects what meal they're eating. Is it breakfast, lunch, or dinner? All three adapt and learn independently. Then they select the carb amount.
Typically, the user just says, "This is my usual amount of carbs for my breakfast." It's approximately in that zone. Or they can select, "It's a lot less than my usual," or, "A lot more than my usual." Those are the three choices. Very simple. They don't need to have precise carb counting. I wish we could show you the actual app, but I'm not sure we're going to have this cooperate today. Wait. Move on. Just give us a bit.
Okay. I'll just pull it up. So here it is. We have the meal announcement button on the bottom. You can select breakfast for this morning. I'm going to have my usual amount of carbs.
It's actually up now, Mike.
Okay. There we go.
Oh, no. It's on the camera.
Hold it very still.
Okay. Here we go. On the bottom here, we just select announce meal, breakfast, usual amount of carbs, and then slide to deliver. My patch just beeped. I do not know if you heard that. Confirming that I have initiated a bolus. There we go. There is my meal announcement. I did not have to figure out how much insulin to do. iLet automatically determines the right amount of insulin for my breakfast as it is learning. That is it.
Thank you, Mike. What I hope you just saw there is a patch pump architecture that really combines the best of both worlds of single-part disposable products and two-part systems. There are obviously benefits and drawbacks to each, and we think we have done a really good job of harnessing those benefits of each one in our particular architecture with the particular design decisions that we have made and that Mike and his team have made.
What that does is it gives us an architecture that's a very compelling patch pump architecture to allow us to access the patch pump half of the market, the people who prefer that. Add that to iLet, which is, of course, a compelling durable product architecture for the other half of the market. That expands our addressable market. Mint is going to be reimbursed exclusively through the pharmaceutical system. Obviously, that's something that we've pioneered on the durable side. We're going to accelerate that on the patch pump side. It's a cost-effective architecture, right? It's scale. By any measure, something where you only have to replace the cartridge, the batteries, and the patch instead of the entirety of all those expensive components is going to be advantaged at any level of scale. As Mike mentioned earlier, environmental waste.
It is somewhat unbelievable how much stuff we throw away in diabetes every three days. We did our level best to help eliminate that problem as well. With that, that sort of concludes our prepared remarks today. We hope we have left some time for questions. Yeah. Right on time, considering we started 10 minutes late. Yeah, we will open it up. Do we have a mic or something? How are we doing that, Blake?
We do, yeah. Blake's going to walk around with the microphone if you guys have questions and output.
All right. Obviously, everybody's available for questions, so dive in.
Hi. Thanks. Rich Newitter from Truist Securities, and thanks for hosting this. Maybe just on the two-piece patch here, can you give us some updated timelines on when you think this could get to the market?
It did not take long to get into that question.
What the process is. And then also, how should we think about the margin impact to you guys and financials? And then if you could also just weave in kind of your go-to-market strategy here.
Yeah. Good question. In terms of timelines, we're not going to update the guidance we've already provided. We're still stating by the end of 2027. Our intention today was to give you an idea of where the product is. There was a particular article that was written that called it, I think it was potentially aspirational at this stage. Probably not. I think you've seen a little bit more output from us than that. Again, the patch Mike's wearing right now is actually pumping saline, and his phone is really controlling it, etc., etc. It gives you some kind of idea. We're not going to get dragged into terms like design freeze or design lock or what have you because they actually do not mean anything. In the actual medical device development world, they just do not.
There are going to be changes to those products all the way up to and post-launch. But what you can see today is that the product is there and exists and is in the form it's going to be at in the commercial launch. In terms of public statements about our launch timing, certainly you'll see a 510(k) clearance when it happens. You know we don't have that because it's a public statement. We'll talk about that when it happens. Don't expect that to be coincident with actual launch timing because, as most people have shared, manufacturing on something like this is the hard part. It truly is. You've probably heard a lot of feedback that manufacturing patches is extremely hard, especially one-pieces. That's true. That's one reason we've gone with the two-piece architecture. Still hard. So yeah, expect.
We, of course, reserve the right to revisit those timelines at the time we make a public statement about our 510(k) clearance, which obviously we haven't done yet. Steven, you want to talk about margin impact?
Sure. The patch was designed with user experience in mind, but it also has the benefit of the design being cost-advantage relative to the other patch on the market. All the expensive componentry is in that two-year durable part, and the much less expensive part, really just a couple of batteries and a syringe cannula, are in the disposable. With the same ASP pricing structure as the current patch on the market, which is what we imagine our pricing will be, at any level of scale, we can manufacture this more cost-effectively. We expect the margin profile to be relative to our patch pump competitor advantage.
Go ahead, Chris.
Yeah. Thanks, Chris Pascual. I had a question about the rainbow charts that you guys showed, which are great, but I look at that breakdown, and it's still two-thirds of the population technically above target. Are there clinical sequelae to being at 7.5 versus 6.5? Why not be more aggressive with the algorithm? Is that where dual hormone comes in? Is that the role you see that playing down the road? Thanks.
Yeah. That's a good question. I think that the algorithm was optimized to get the best clinical results that we could get while having minimal levels of hypoglycemia. You would definitely be having greater risk of hypoglycemia if you try and push harder, especially in some people who've had much more challenges with controlling their blood glucose, may not be announcing meals. I will say that for people with A1Cs below 7, you probably saw on the chart that the A1C to GMI goes up. We thought that that might be because of some peculiarities of A1C down in the low range. We have done an analysis of people for whom we had follow-up A1Cs. Whereas we go from 6.4 to 6.9 or 7 in most of those rainbow graphs, if we go A1C to GMI. If we go A1C to A1C, it's 6.4 to 6.5.
We're able to maintain glucose control for people who already have good glucose control well below 7. I think one of the main differences there is lifestyle. If you have a lower-carb diet, it's much easier to have tight glycemic control. If you exercise, if you do a number of things that can help your glycemic control, you get better glucose control. What we see is many people who come and talk to me at this conference who had excellent glucose control on some other system, they come into the iLet, they get even better glucose control. I think it's not really a limitation of the system, except that we're trying to be careful with not causing hypoglycemia. Can you repeat the second part of your question, please?
The dual hormone.
The role of dual hormone. Yes. That's an important point. What we saw in our pre-pivotal studies with dual hormone is that we got about an additional 0.5% of A1C. In the pre-pivotal studies, which were a less challenging cohort, admittedly, than our post-market, we were seeing about half of the people in those pre-pivotal studies be able to get to a GMI below 7. That's a little higher than what we're seeing post-market. As a point of comparison, in those previous pre-pivotal studies, in that same cohort, we were able to get 90% of people to a GMI below 7 with the bi-hormonal system. We shifted that whole distribution downwards. There was only a little tail above 7, and all of those wound up between 7 and 7.5. That's the power of bi-hormonal. It allows us to shift that entire population downwards.
At least in those pre-pivotal studies, we also saw further reduction in hypoglycemia combined with that. That is the promise of being able to go with bi-hormonal.
Yeah. Chris, it's important to note that the DCCT taught us that there's a reason 7% is goal, right? You meaningfully eliminate complications as you approach 7% A1C. As you go below that, you start to see hypo increase, as he mentioned. There's a reason that we shoot for 7 and not 5. So there you go. With bi-hormone.
Yeah. I have one other point about that. That is the DCCT found that there was kind of a hockey stick-shaped relation of complications. Above 7.2 or so, it started to go up. Keep in mind, those people on average started with an A1C of 9, and you can't remove that damage that has already occurred. There are some interesting observational studies from some Scandinavian investigators where they're able to follow people from the time they're diagnosed throughout their entire lifespan with diabetes. They found that if, I think I'm going to get this right, if they got to an A1C below 7.5 within a year of diagnosis, there was no signal for complications. It may be that if you get good glucose control early, you might not need to get all the way down to 7. Maybe 7.5 would be good.
Another piece of evidence is there's a small group of people, several thousand people in the world, who have glucokinase mutations. They start off from birth having a much higher glucose, an A1C between 7 and 7.5. They stay that way throughout their entire lives. They have normal lifespan, and they do not get any complications of diabetes. I personally think maybe 7.5 is good enough to dramatically reduce complications in the modern environment.
Thank you. Good morning. Jeff Johnson from Baird. Sean, two questions for you, I guess, on pipeline beyond what you've talked about today. You guys filed a T1, T2 combined study in the last week or two on clinicaltrials.gov. Should we read that as being your T2 label expansion study? Would that study be powered enough on T2 to potentially pursue label expansion there?
Yeah. You want to take this one?
That study is a primary care study. You may have seen that there was a study published by Sean and Tamara Ozer, our colleagues at the University of Colorado. They did a preliminary study where you saw some real-world data comparing the endocrinology office and the primary care office. That was a study where they in primary care recruited people who were MDI folks who did not have experience with pumps, and then they put them on the iLet. I started this study when I was still full-time at Mass General Hospital. It was carried on by Melissa Putman, my colleague, where we recruited people who were pumpers already and put them on the iLet. What we found is that the outcomes were basically the same. It was sort of similar to the real-world data.
Now, Sean and Tamara Ozer are going to expand on that study. It's funded by the Helmsley Trust, and they're going to recruit a much larger number of people out of the primary care setting by primary care doctors. They're supervising the study, but they're not the ones who are going to be recruiting them. They're going to be testing their performance with using the iLet. That study is recruiting both type 1 and type 2, partly because sometimes in the primary care world, it's not entirely clear if people are type 1 or type 2. We wanted to make sure to include both. However, that study is not powered, we think, to be able to do an indication, to get an indication, and to expand the label.
However, it will give us a lot of important information about how a study like that would have to be powered.
All right. Great. Any updated timeline then on when you might pursue type 2? And you showed a nice little chart there of iLet kind of moving towards fully closed loop. What's it take to get there? And do we need to see a number of pre-pivotals before you'd move into a pivotal? Are we talking a 2027, 2028 closed loop iLet, or how to think about that? Thanks.
No updated timelines. I may need to ask you to prompt me in the second half of that question in a minute. I want to reiterate, though, on type 2 that in the Q1 earnings call, we stated that, make sure I get this number right. Stephen, is it roughly 25? What did we say?
20%-25%.
20%-25% of our users are coming to us with type 2 diabetes. I think it's important to note that it's more difficult, not impossible, to get a pump approved for type 2 because of the traditional negative C-peptide requirements. Now, in pharmacy, that tends to go away. In our case, we also released that we were whenever it was low 20s% pharmacy in Q1. If you think about that on a per-covered patient basis, we're actually quite well utilized in type 2 without the label. It's just a way to think about that. No updated timeline on that. I'll leave it alone, but just wanted to provide additional color. The second half of your question was fully closed-loop iLet. What does it take to get there? We don't know. Nobody's ever had that indication before.
What we showed today is the fact that iLet is being used by a subsegment of our patients. We do not know self-selected or how that works, but a large percentage, 16%, in a very acceptable way. Does that mean submit that data? Does that mean do a study on fully closed loop with iLet? Does that mean optimize iLet for fully closed loop with some other change? Any of those things are available to us. We do not know which one will necessarily be acceptable to the agency. We are going to learn. Excuse me, as you will.
I want to add something.
We're certainly thinking about it.
I wanted to add something on the type 2. We do have a poster on the outcomes in our real-world data divided out among type 1 and type 2 that's available in the poster hall.
Thanks. Matt O'Brien, Piper Sandler. Two questions. I'll just do that first one up front here. The acceleration in the pharmacy for the year, maybe this question's for Stephen, what drove that quick change versus what you guys had originally laid out? Can we associate that bump due to an acceleration in the HCPs that you're adding because of the ease of use? Maybe Stephen, just talk a little bit about the economic benefit that you see in year two and three from that versus maybe what we had initially been expecting here in year one.
Yeah. Sure. Mark, are you comfortable taking that?
I'll take the first part and take the financial part. The increase in pharmacy has most notably been driven by adoption, and it's been augmented by our increased efficiency at getting the appropriate patient down the appropriate channel. That is why you have seen the numbers increase. As I mentioned in my prepared remarks, I had the historical experience of seeing how adoption occurred at Dexcom, and things have definitely gone faster here than they did from approximately mid-2014 through mid-2016 back with Dexcom. We have been pleased with that progression. We raised our guidance to 22%-25% of all new patient starts for this year.
Mark, I just want to be super clear on that. When you say adoption, you mean adoption by the plans, not by patients or healthcare providers, right?
Adoption by the plans, yep.
Yeah. So briefly, a reminder of how the pharmacy business model actually works. We're contrasting a DME. In DME, we get a large upfront. We recognize a large upfront amount of revenue when we sell the pump, so roughly $3,500 of revenue on day one when we sell the pump in DME, and then about $70 a month to sell the supply kits. Contrasting that to pharmacy, we recognize around $0 upfront when we sell the pump, and then $450 a month to sell the exact same kit of supplies. Doing that math, where it's financially advantageous to Beta Bionics is after 11 months, and it's about 3x the amount of reimbursement over a four-year period. Every patient we can send to that channel, every patient that we can send to pharmacy, we do, because to Matt's point, it's advantage to us financially in the medium and long term.
What actually you see in our financials, if you look at, say, Q4 2024 to Q1 2025, you saw a gross margin reduction. The real reason for that, even though the revenue was like it was only a small reduction from Q4 to Q1, was that we saw a big uptick in pharmacy adoption. We went from a low teens percentage of our business, of our new patient starts in Q4 going through pharmacy, to a low 20s percentage in the first quarter. That creates some short-term headwind, but that is a very, very lucrative, powerful install base in pharmacy that now has a huge benefit for us in future quarters. Again, after month 11, assuming that the patients stay with the product, which we have a lot of confidence in and we are seeing in the data, it is a much better business model for us.
Got it. The follow-up is for Sean. Are you to the point, and sorry, it's another patch question, but are you to the point now where you're corresponding with FDA on human factors specifically, or is that something to come? You have to still kind of finalize some things on the development side. How difficult is the manufacturing process? We're not going to see like a year lag, or are we like, "Hey, we got approval in early 2027, but we can't make it until the end of the year." Thanks.
First of all, I admire your ability to get a little bit more information out of me on this. I did not think that was going to happen. Nice work. Look, FDA interactions are constant on products from way before you even start the product all through until you are cleared and, in fact, after as well. So have we talked to the FDA on this product? Yeah, of course we have. I am not going to give you any additional benefit on what we have said to them and what interaction we are in, et cetera. Yeah, of course, this would not be a surprise to them. Let us put it that way extremely generally. And then on what was the second one?
The second part is just on the manufacturing process.
Right, manufacturing process. I'm not going to comment on that today, but I'll reiterate that at the time, the 510(k) clearance, we'll be in a better position to know what it's going to still take. It just depends on, obviously, when that clearance would ultimately come and where we are in the manufacturing process. Let me say it in another way. It doesn't take 12 months to build a manufacturing line for a patch. It takes a hell of a lot longer than that. All right. That makes it extremely important as to when we get that clearance.
Thanks.
Hi, good morning. Jakob Dodd, Morgan Stanley. I found the discussion on the adaptive closed-loop really insightful and compelling, but I'd be curious to know how has Beta been able to achieve that level of adaptiveness, whereas perhaps some competitors have not over the years? And how durable is that algorithmic advantage going forward? Thanks.
Yeah. That's a really good question. I think there's a bunch of answers to it, and I'm also interested in Stephen's perspective, but let me take a first crack at it. The way we've developed a product is very different than the way everybody else developed theirs. That's just sort of—it wasn't any great insight. It was just sort of the way it happened. Traditional pump companies started with, we'll call them dumb pumps, and I don't mean that to be dismissive. They're great products, developed in the 1980s, and they moved forward, and we added features on top of them. A lot of those features are going to be mentioned today, and Mike mentioned extended boluses and square wave and dual wave and all this stuff, features that have been added.
Hybrid closed-loop systems were features that were added on top of traditional insulin pumps, insulin pumps that already had settings in them and all those requirements that we talked about today that use that math that I talked about today. They were developed in that light. iLet was done completely differently. In fact, the algorithms that iLet is based on were developed before the founder of iLet, Ed, and our clinical collaborator, Stephen, had any pump to work with. They were run on a laptop. They were not confined to the thinking of what a traditional pump did. They just asked themselves, "What should it do?" In fact, when you think about it in terms of how hard it is, when did that work occur?
Steven can double-check me, but I think the first animal studies done on this product were done in, actually, I do not even know that. First human 2007?
First human studies 2008.
2008. First ambulatory study 2013. Sorry, ambulatory out of the CRC trial.
Got it.
We can help.
Yeah, 2012.
When you think about it, the first study that was done that way on any hybrid closed-loop system was actually done by us. We have been at it a really long time. It is actually hard. I showed you the basic math on how dosing works every five minutes. We did not get into the details of each individual algorithm. I said glucose is high. Subtract target. That is how high you are divided by your insulin sensitivity factor, right? Great. A correction controller in a hybrid closed-loop system might include a rate of change term. It is not really all that much complexity on top of those existing dosing algorithms. How the hell do you learn that ISF? In terms of the spaghetti chart that I showed you, that is the hard part.
At some level, you're just going to have to take my word for it. The adaptive portions of the algorithms are hard. The real-time controller, which is that math that I showed, is comparatively easy. I'm not trying to take anything away from it. They're still hard. There you go. Yeah, we do believe there's a durable competitive advantage here. Again, you have seen people start to Markh toward adaptivity, but still pretty far away from where we are.
Great. Thanks. Steven and Brandon?
Yeah. I would just add that we were kind of forced to make the algorithms adaptive because we wanted to have a really different user experience. We wanted to eliminate carb counting. We wanted to eliminate the need to do manual corrections. Those design decisions were made a very long time ago. We realized, if that's going to be the case, it's going to have to have a robust autonomous adaptation. We iterated that through a lot of clinical trials. We did 20 pre-pivotal clinical trials every time iterating. We learned a lot from that process. I think going forward, we can move a lot more quickly than that to further improve our algorithms. I think what we and moving to closed loop, we're not having to make two jumps.
We're not having to make the diagonal jump from a static system to also a fully closed loop system. We already have the adaptive nature of the device, and it already works really well in closed loop, a fully closed loop for a lot of people. We just have to make that lateral move to make it work in fully closed loop for everyone, essentially.
Thanks.
Hi. Matt Miksic from Barclays. Thanks for taking the question. Just to follow up maybe on Chris's question around the destination where you're landing slightly above seven, a couple of things. Is there an analysis of that where you're able to sort of take out the folks who are not putting in any meal investments to give a sort of non-hands-off profile of what those patients look like? The second part is anything about, I don't want to call it gamification, but encouragement of folks who aren't to get incrementally more active to kind of show them that they can nudge that number down a bit. Thanks.
That's a really interesting question that I guess I'll be honest to say, I've never really thought of. I'll tell you why I haven't thought of it. If I understand your question correctly, what you're asking for is subsegmentation of the data, including only people who are meal announcing as requested. Let's say average of three per day. Fair. We haven't done that because we don't tend to subsegment our data in ways that benefits it. We tend to look at subsegments that are tougher. There's been a, I think, lamentable tendency in our industry to do sort of exactly that. It's caused a situation where you have had a very hard time getting through every individual clinical trial and saying, "Okay, that looks great. Why?" We'd rather not do that.
We'd rather use the thing in a broadly applicable trial and just tell you how it did in the real closest to the real world. We believe that our pivotal trial was as close to predictive of the real world, and I think that's proved true as anything that's been done. My pain is there you go. That being said, I don't know what it would show because I haven't looked at it. Better. If I remember correctly, in I'm not sure if it's our real-world evidence or a clinical trial, we have an average number of meal announcements of about three. Can you help me out here?
Yeah, it's about. We have looked in our one-year real-world evidence. We haven't done it for a two-year yet where we looked at outcomes according to number of meal announcements a day. We have done that. For people announcing, I think it's three or more meal announcements per day, the GMI is 7.1, I think. It is lower. What is remarkable is that people in that group who are announcing only one meal every three days are getting 7.5. To us, it's more remarkable that the difference is as small as it is. The other thing that I totally agree with, Sean, I think you'll see from competitors, people using the recommended settings, the lowest target, the shortest insulin action time. They're encouraging people to use more aggressive correction factors and sensitivity factors.
They like to report just the data for the people using the most aggressive factors. That is not real world. It is a small percentage of their populations. We have always felt that we need to take people as they come. We know that populations that announce less meals on the iLet have higher baseline A1Cs. That amount of meal announcements may just be the amount of attention they can give to their diabetes. They still deserve to not die early and with complications. We are focusing on trying to get the best outcomes for everybody, regardless if they are willing to use the optimal settings and they are willing to be as safe as the device, essentially. We want to work for them rather than the other way around.
I think it's a perfectly valid goal to report and try to get the best latency time range, I don't care, for a certain subsegment of patients, a nearer group. That's great for those patients. I think we've laid out today, and both these guys talked about, is our goal is completely different. That's to improve the population health. We're not going to do that by focusing on the uber-engaged patient and getting them to an A1C of six or whatever it is. We're going to do that by focusing on the entirety of the population and getting them to target. It is a controlled-to-target system. Target is seven. In our clinical trial, 48% of people got to seven. Center of the bell curve. We nailed it. I think we're at time. I guess, Taylor, do you have one more final thing? Sure.
Yeah. I just wanted to know your response to that question about the over 12 months to build a live firm patch.
Yeah.
Just trying to understand, in the late 2027 goal, do you need five to eight approval at least a year before that? Or are you hoping to build into the line in parallel and just separately, just to any capacity at launch at the end of 2027? What can we expect?
Yeah, good question. I'll take half that question. I'm not going to talk about what our capacity will be at launch. What I probably will say is we're not going to be looking at a massively limited launch. We'll do a broad launch at the time we're ready. I'll additionally say that we're not waiting for a 510 clearance to kick off the manufacturing line. Just like we always talk about, we have confidence that users like our product, that they're going to stay in our product, that we're not worried about things like attrition. We also have confidence that our technical team and our regulatory team are going to get this product cleared. We're not going to wait for that. I'm not going to go any farther than that.
If you guys are interested in seeing the patch, by the way, after, Mike will be around for a while. If you want to see the app in person, Michael will show you after. Thanks, everybody.
All right. Thanks very much for coming. Really appreciate it.