HeartFlow, Inc. (HTFL) Earnings Call Transcript & Summary

All right. Thanks for being here, everyone. Great to see you. Glad to have the HeartFlow team with us today. Maybe quickly before I jump in, you got to grab your research disclosures from our website at morganstanley.com/researchdisclosures. If you have any questions, please reach out to your Morgan Stanley sales representative. So thanks for being here. Really excited to have HeartFlow with us and dig into another health care AI story.

Maybe, John, just to start, if you could just sort of hit where you guys sit in the market, what your application is, what the need is and sort of what your value prop is?

Yes, sure. So thanks for having me. Pleasure to be here. So we're a technology company. We use AI. We deploy AI to better diagnose coronary artery disease. So at our core, that's what we do. I think there's a tech conference, but you're probably all aware of heart disease in and of itself is the world's #1 killer. So we think we're deploying our technology to a very serious problem. It kills more people than all forms of cancer combined. Every 40 seconds, somebody in this country is having a heart attack, half of all those heart attacks are a total surprise, meaning they've been through -- the patient has been through the health care system, and it hasn't appropriately been diagnosed. So that's really the problem that we're trying to solve, and we're in the very early innings, but we believe we can create a new standard of care using our technology to do that. And everything we do is backed with clinical data. Our clinical data proves our accuracy. It proves our outcomes. We're very proud of that. What's unique in our AI and the insights that we deliver to our physicians is the use of our technology is reimbursed. Well, first off, it's all cleared by the FDA. But it's reimbursed. So it's covered by Medicare, it's covered by commercial payers. So our physician customers and their hospitals, they make a margin by using our technology. And similarly, we have separate revenue streams for those reimbursed technologies as well. So we went public in August of -- just this last August and super excited for what the future holds with our technology.

Maybe we can spend a second sort of double-clicking on the noninvasive testing market for cardiology. What does that look like? maybe spend a second on CCTA, which is the underlying imaging modality and whatnot.

Yes, sure. And I should mention our technology sits on top of a coronary CT image, okay? So that's what we deploy our AI to. But...

You don't own that...

We don't own that. We deploy that into our cloud, and we put our technology on top of that. Now relative to the -- as you referred to it, the noninvasive testing market, if you take a step back, our current patient population that we're treating with our technology are patients with symptoms, okay? So that's -- you're mowing the lawn on the weekend, you're running to catch a flight, you're chasing your grandkids, what have you, you have some type of a symptom. That could be chest pain, shortness of breath, dizziness, fatigue. Ultimately, that patient finds their way in most cases, to a cardiologist. And that cardiologists job is not to run them straight into the hospital and do an invasive procedure, is to diagnose those symptoms noninvasively, okay? So that's the market that we're sort of playing in currently. In total, there's 9.5 million noninvasive tests every year in this country. If you monetize that against our technology, that's about a $5 billion TAM, okay? Now our penetration within that is just like I mentioned, the really early innings, we're about a 2% penetration against that total noninvasive testing market. But again, we believe we're on the right side of history here, and we can create a new standard of care. You asked about the coronary CTA. So the coronary CTA, again, is what we -- that's the frontline test that patients in some cases, but in an increasingly amount of patients are getting ordered for this noninvasive test. That's about 10% penetrated, and then we're a fraction of that.

Maybe just spend a second actually, what are the other noninvasive types of tests?

Yes. So these are stress tests, treadmill tests, sometimes you call them nuclear tests. And the vast majority of the market still relies on those. And we really believe our technology is better and can improve on the inaccuracies. Just a note on those, and this is well understood in the clinical data. 55% of the time, they diagnose with a false positive, meaning you'll go and you'll get a nuclear test and be sent into the cath lab when you don't need to be there for an intervention. More concerning, 30% of the time, they miss it altogether. And that's when these patients go home and unfortunately, have have bad events at home. Again, we have clinical data that proves we're much more accurate than that, and we can avoid those false positives and those false negatives.

So patient ends up with the cardiologist, gets a CCTA, image goes in the cloud to HeartFlow. What's the platform and the products and solutions that come off of that then at that point?

Yes, good. So first of all, our -- we have a technology platform. We're not a singular test, okay? And our platform, and we're committed to this, but our platform continues to expand year-on-year, okay? Right now, we have 4 components to it. The first sits at the front end of the workflow. And I should say wherever that patient is in their journey, whether or not it's detecting it, diagnosing it, managing it or treating it, our goal is we bring clinically validated, reimbursed insights to our clinicians, so they can take better care of their patients, okay? The first part of our platform, we call it the road map analysis. This is a workflow tool. We have clinical data that substantiate -- it allows CT readers. So these are the physicians that are reading that CT image to read it faster with less variability, okay? And we provide that free of charge for every CT that's sent to HeartFlow, okay? And it's a really nice component of the model, and I can speak more as to sort of how it enables some other things in a moment. Next, sort of in between the bookends here, we have 2 reimbursed technologies. Each serve a separate insight for clinicians and each are reimbursed separately. So there's separate revenue streams for us. The first is our plaque analysis. plaque analysis measures plaque, whether it's calcified, noncalcified or low attenuation plaque down to the cubic millimeter. And so that's orders of magnitude more precise than any physician could do with the human eye. And again, we back this with clinical data, so we know it's accurate against the invasive gold standard, okay? Our next reimbursed piece of the platform is our FFRCT analysis. What FFRCT does is it models the blood flow across any disease that's found and tells the physician whether or not the blood flow has been impeded to the point where that lesion or that disease spot of the coronary tree needs to be intervened on, okay? So separate insight than plaque and again, separate reimbursement for the physician. And then lastly, which we're launching here shortly this year, we have what's called a navigator tool, a PCI navigator tool. So again, if you follow that patient all the way through the continuum and they end up needing to be in the cath lab with a disease that needs to be treated, we now have a tool where the interventional cardiologists can plan accordingly to take care of that patient. So that's a platform as it stands now. But again, we fully intend to continue to add components to it.

So maybe a couple of other questions on the platform. One of the things I don't think we've talked about yet is one of the unique, I think, advantages, although maybe it needs a little explanation is you've got a human in the loop still on your side. So not just the physician treating the patient, but on your side, you've got a human in the loop. Maybe just talk about how that works, John, why that is or isn't necessary? Could that human go down to 0 over time? Can it not, et cetera?

Yes. So first thing I'll say, again, we're in health care. This is regulated. This is FDA-cleared technology. This is technology that needs to operate within a regulated environment. As part of our FDA clearance, we are required to have a human in the loop, okay? That human serves a really important component. They ensure quality of the algorithm, okay? So the first step that occurs when we ingest these CT images from hospitals and clinics, it comes to our cloud. And the first thing the algorithms do is they create a precise 3D model of that actual patient's coronary tree, okay? If that model is perfect, then all of our algorithms can run autonomously off of it, okay? But because CT images come and sometimes they have artifacts in them, sometimes they might be blurry based on a patient's heart rate while they're getting the scan, when they show up, they're not always pristine. That's where our human in the loop, we call them a quality control analyst make a couple of adjustments all the way, okay? And the analogy I sometimes use is if you have a Waymo taxi driving around San Francisco, if you give it a perfect map of the city, they won't run the stop sign, they won't drive up or they won't drive up on the sidewalk, what have you. Similar here, what these humans in the loops is they're making that city map perfect and pristine. From there, the algorithms take over, and that's how our technologies are sort of brought to fruition. Now if we were having this conversation, you asked kind of how long does it take to do. If we were having this conversation 15 years ago when the company was founded, that quality control analyst probably spent 20 hours doing those connections, okay? Last year, at the end of last year, they spent about 20 minutes, okay? So every...

Has that fundamentally have been a technological improvement?

Yes, exactly. It's 100% what it's been. So every unit of time that's taken out is a function of our algorithm getting smarter over time as we feed more data. So we've taken a lot of time out over the years, and you can look at our margin expansion over the same period of time and see that occurring. And I've got super high confidence that, that expansion is going to continue in the future. To get to your last question, will there always -- can we always take a human out altogether? We never say never. I think where I've got really good line of sight on is we're going to take that time down and down and down. If you were to take the algorithm or the human out altogether, that's a new regulatory filing. And we'd have to evaluate whether or not that's a strong business case when it's appropriate to.

So you -- maybe just spend a second on the algorithm. You've talked about -- you've got 150 million-plus annotated images. You talked about the algorithm improvement over time. There's a lot of talk at the conference and the investor community right now on, how proprietary or not is the data, how proprietary or not is an algorithm or piece of software. Maybe just talk us a little bit about the training and building of the algorithm, the testing that's went into it, the proprietary nature or not of the data you have, et cetera.

Yes. So good question. So first thing I'll say is, right now, we've got, give or take, call it, 116 million annotated CT images, okay? So these are clean CT images that have been, for lack of a better word, kind of scrubbed by our analysts over the course of time. This is a data set that's incredibly diverse. We've been growing it for 15 years. So as different anatomies come through the cloud, as different types of CT capital comes through, we've got a very diverse data set there. This data set, to the best of our knowledge, is the largest that exists, and it is proprietary to us. And like I said, we've grown it over time. And as we speak, it's growing, okay? Every time we ingest a CT, we grow it. So that's one really important kind of component of the business model. And we use that data to create new algorithms to expand gross margins. We use that same data to create new technology that we launch back to our existing customers through those same data pipe and connections. So that's one really important piece that I think serves as a compelling moat, so to speak, as we move forward. The other is around -- as I mentioned, this is a regulated space, okay? You need FDA clearance. You need to know how to approach the FDA in order to get that clearance. You need to know how to produce products at scale and deliver them back to physician customers underneath a quality control system that's auditable by the FDA. You need complaint handling, you need design controls. No different if you were an implantable device. So that's a skill set that we've built, and I think we're pretty good at over time. There's another kind of data that's, I would argue, equally as important as the data we use to train the algorithm, and that's our clinical data, okay? At the end of the day, our customers are physicians. okay? Physicians, by and large, make decisions on how is this going to help me treat my patient better. The way you impact practice patterns and change behaviors. So remember, we're trying to shift behavior from this traditional standard of care to this new one is through strong clinical data. And the company has invested over time. We were founded in 2010 and credit to the team in 2010 for making the commitment to do this. We have over 600 peer-reviewed clinical publications in the market right now. That's orders -- and that's taken us 15 years to accumulate. That's orders of magnitude more than any other -- anybody else has got. And we're continuing to invest in that. And then I think the last piece is we work really closely with customers. And again, we've learned this over the years on how do you take this technology and integrate it into a health systems ecosystem, okay? So these health systems, there's a lot of different stakeholders. You could be an interventional cardiologist, you could be a CCTA reader, you could be a general cardiologist, you could be an internist or a primary care physician, okay? And after we kind of go live in an account, we work to integrate across that entire ecosystem. Now that can be challenging in a lot of respects, and it's taken us a long time on how to do it well. But where we are now, we really think it's another kind of compelling piece of our moat because once you're in, the switching costs are pretty high, and we see a lot of really kind of sticky behavior. And then the last piece, I should say, everything we do is backed by really strong IP, and we've got a good IP portfolio as well.

You mentioned the sort of bidirectional data pipes. I think that's actually an exceedingly rare asset or capability across health care. I talked to Tempus earlier today, they've got that. But if you go talk to a very large percentage of our clients, they don't really have that in any form or fashion. Maybe they're passing a PDF to their client or something, but there's not a true data relationship. Can you talk a little bit about what that looks like for you, what it's been like to build that, where you're at in building those connections and where you need to go sort of in the future?

Yes. Okay. So first thing, just for kind of context here. At the start of this year in the U.S., we were connected to about 1,500 accounts roughly, okay? And those are hospitals or those are clinics, okay? Anyone that has a CTA within the building. Once we're into those accounts, that's where this bidirectional data pipe comes in. We're in about 1,500. That number again is growing just about every day as we're adding more. The total hunting ground is around 3,200 accounts. And that 3,200 is growing by about 300 a year, okay? And I just say that to say there's a lot more pipes for us to kind of connect. And if there's 300 new ones entering the category every year, that's going to keep us plenty busy.

And this is just U.S.?

Yes. This is just U.S. So it's by virtue of that, that we set up these data pipes. Now once that occurs, the minute one of these patients that I mentioned, the patients coming in for a noninvasive test for -- to diagnose coronary artery disease. The minute that patient has a CCTA, it comes to our cloud automatically. Once it's in our cloud, we digest the data, we create the analysis and we bring that analysis back. So by the time the physician is ready to make a decision or make a diagnosis, our technology is there waiting for them, okay? The benefits that the customers yield or glean from this model is they get instant access to our technology the minute they need it. They get access to our road map technology for every one of their CCTAs that's irrespective of whether or not they need any of our other tests. And they get integration into their workflow, okay? So those are benefits for them. From our standpoint, there's a number of benefits as well. We get to consume that data. It builds into our database. So that 160 million annotated CT images that's growing and growing and growing. We mine that data to create new algorithms. Those new algorithms help us expand our gross margins. Those new algorithms help us create new technologies. So as we launch new components of our platform over time, we have that data. And then because we have those data pipes, as we launch that technology, we do it in a very frictionless way, okay? It literally is a new button on their already existing user interface. So physicians like that because they get it quickly and easily. And obviously, we like that because it's a very efficient way to bring new technology to market.

And I would assume there's at least some form or fashion of a moat there because you went through the Infosec reviews and all the things at the hospitals.

Yes. So in order to get -- great point. We've got really high confidence that we've got a good model where we can go get the next new account. And we've done that to the tune of 1,500 accounts currently. That process, however, you need to go through -- first, you need to build clinical champions. But then beyond that, you've got to bring in administration, you've got to bring in IT, you got to bring in information security. In some cases now, you've got to kind of navigate a sort of an AI board at the health system. That process to land that next new account can take a lot boards every year, okay? So again, a little bit of a barrier going in. But once you're in, it makes for a pretty sticky customer on the other end.

Maybe we could double-click on one of the things you mentioned, which is plaque. That's a newer sort of offering, at least from a commercial perspective for you. Maybe talk a little bit more about what that is how that launch is going, sort of what you expect out of that? How important is that going to be over the next couple of years?

Yes, perfect. So just for reference for the audience, the vast majority of our current revenue is all coming off our initial test, that's our FFRCT test. And that's a durable business that we believe is growing very nicely with lots of upside relative to penetration. The next kind of layer of cake, so to speak, of growth is our Plaque Analysis. And this is an analysis that's FDA cleared, and it's paid for, both through Medicare and a large growing number of commercial payers. And we started launching this last year. And at the start of this year, we're in close to 500 accounts just at the start of early January. We think we'll be north of 1,000 by the end of this year, okay? So we're really excited about.

Is that largely your existing account base you're going and penetrating?

The higher the majority -- the majority of that is existing, but there's some news as well. Yes, good question. So we're super excited about kind of the market interest on plaque. And plaque from a patient applicability standpoint, has doubled the applicability of FFRCT. So more patients can benefit from this reimbursed technology, okay? Now we're still in the early innings of that. So there's a lot of kind of the initial phases of this is you got to sign a contract, you got to take them live. Now we're in this -- help them ramp and use it. Now there's a lot of questions out there on exactly what is the right way to use this analysis. That's probably the #1 question that I get that the team gets. So we're leaning into quite a bit of medical education to help physicians understand how to use it. Now that being said, based on all the early indicators, I'm extremely excited about what 2026 has for plaque. And I think by the second half of this year, we'll see some pretty material impact by virtue of all the early demand we've seen.

You mentioned how much it expands the patient population. Could you spend another second on who is plaque applicable for? What's the exact?

Yes. So it's covered. The applicable patient population is really driven by the coverage that the physician will get paid for it. Where FFRCT is relevant to about 1/3 of all patients based on their disease burden, plaque is about 2/3. So much bigger. And so you could go get a heart flow analysis, not need an FFRCT, but very likely you're going to need a plaque. And I should say with plaque, we're measuring disease down to the cubic millimeter. And a lot of people probably have had calcium scores. That's usually a question I get. Calcified plaque is one type of plaque, okay? There's noncalcified plaque. There's low attenuation plaque. Physicians really need to understand the full profile of what your composition of your plaque burden is in order to better treat you. And that's what we aim to do with our analysis.

Right. Much higher fidelity set of information around...

Yes, you can't get it with the human eyeball, and it's not appropriate. There's no physician out there that's going to send you into the cath lab for an invasive procedure just to understand your plaque burden that wouldn't happen.

Which is effectively the gold standard. Invasive is the gold standard. Right. Okay. So we talked about sort of plaque, how that could impact the near term. It sounds like you've got a lot of excitement there. As you maybe try to look out the next several years, what is it that excites you guys? What are you sort of focused on? What are the growth and value drivers as you play forward here?

Yes. So I think in the whatever you want to call it, near midterm, super healthy kind of base business. Our core FFRCT business has tons of runway ahead of it. And that's the beauty of creating a new category is the more we're shifting towards CT plus HeartFlow, the more runway there is. So there's great opportunity with that in and of itself. Plaque coming online is super compelling and the top focus of the teams right now. But we still have this database, right? And we have a very long track record of using that database to introduce new technologies that we then can launch through the same bidirectional data pipes as we call them. In 2026, the next element that we're launching is our Navigator. This is our PCI Navigator tool. This, again, is intended for downstream usage. This is interventional cardiologists, and they can use our platform to help plan for a PCI procedure, okay? We really like this for a couple of reasons. One, that interventional cardiologist can be a great champion for HeartFlow within the health system. So if they use our technology on all of their patients, they're going to help champion a CT-first pathway across their health system. So that's a compelling piece. The other piece is we believe in the power of the platform. The stronger the platform, the stronger the platform. And we know we can deliver value to an important stakeholder through this. So that's our '26 kind of platform expansion. In '27, we have a what we call a serial plaque analysis planned. And so that's -- you can take -- where you can track disease progression or hopefully regression from Scan A against Scan B. So if you think about this longitudinally, 2026 is really the year of patients getting their first Plaque Analysis, okay? You don't need a plaque analysis every year, okay? But then as you go into '27, and you're working with your cardiologist and you want to understand whether or not the therapeutics that you've been on are actually impacting your disease burden, what do you need? You need your second scan, okay? So we're going to launch a serial plaque technology that will co-register against the first and allow physicians to see exactly how the disease burden is being changed, hopefully in a positive way relative to the therapeutics that, that patient has been on. So that takes us through 2027. Beyond that, we have other technologies planned, but not yet for disclosed discussions. We're very excited about opening up a new addressable market, okay? So everything we've been doing up until now, has been symptomatic patients, patients with chest pain, that's a $6 -- excuse me, $5 billion total addressable market. The nextdoor neighbor to these patients are high-risk asymptomatic patients. okay? And then beyond that, you have sort of at-risk and low risk. We're going to enter the high-risk asymptomatic market next. That's another $6 billion TAM, okay? We have high confidence that our existing technology doesn't discriminate whether or not you're having symptoms or not having symptoms. What we need is we need clinical data in order to open up that market. And so we have randomized controlled trials that are planned this year against 3 subpopulations. One is patients with high calcium scores. The other patients are patients with prior heart attack. And then the third is patients with prior plaque. And we'll have separate clinical investments against each of those subpopulations to open up those that high-risk asymptomatic market. All of that, we believe, is going to hit us or help us before the end of this decade.

What do you -- what do those trials fundamentally look at? What do you have to demonstrate to get FDA clearance or to get adoption from the market?

The 2 primary endpoints are going to be plaque regression. So again, how much does plaque change over time? And then second, LDL. It's change in biomarkers. We'll capture -- sorry interrupt. We'll capture longer-term hard outcomes, but we don't believe that's needed to open it up.

And just maybe one last question on that. You mentioned now a couple of times sort of plaque regression or comparative plaque analysis effectively. Does that over time start to get you more integrated with the pharmaceutical companies? Like is there a deeper partnership there in the way that they're delivering drugs and you're measuring success?

I think the opportunity is absolutely there. And what we are able to do is take a picture to tell whether or not the drugs are working. We'll stop. So the goal of these drugs is to prevent the disease from growing. And again, there's a lot of analogies here between what we do and how cancer has evolved over time. But if you find a tumor, you treat it. and you take another picture of the tumor and you make sure it's getting smaller or it's not spreading. That paradigm doesn't yet exist in coronary artery disease. But we believe with our technology, we can get there.

Great. Awesome. Well, thanks again for coming, John. Really appreciate it. Exciting year ahead for the HeartFlow team.

Yes, great. Thanks for having me.