June 18, 2025 • 45 mins
“We started Cleerly with the intent to make a comprehensive care pathway for evaluation, education, treatment and tracking for heart disease,” said Dr. Jim Min, the company’s founder and CEO. In this Vanguards of Health Care episode, Min sits down with BI analyst Matt Henriksson to discuss Cleerly and its comprehensive management platform that uses AI algorithms to boost the accuracy of interpreting CT scans, avoid false positives and make the diagnosis stage of treating the disease more efficient for doctors and hospitals. They also cover the need to drive the iterative process of clinical studies, including the TRANSFORM randomized trial for screening asymptomatic individuals with the aim of preventing sudden cardiac death.
See omnystudio.com/listener for privacy information.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:18):
Welcome to another episode of the Vanguards of Healthcare series.
My name is Matt Hendrickson, the medical technology analyst at
Bloomberg Intelligence, which is the in house equity research platform
of Bloomberg Elpeep. We're pleased to have with us today
doctor Jim Mint, founder and CEO of Clearly, a privately
held medical device company that is applying artificial intelligence to
(00:40):
non invasively measure plaque stenosis in the likelihood of a
schemia within the heart. Jim, thank you for joining us today.
Speaker 2 (00:49):
Thank you for having us man like, it's a pleasure
to be here.
Speaker 1 (00:52):
Likewise, and why don't we just start with an overview
of your career path and then the steps in that
path that led you to founding Clearly.
Speaker 2 (01:02):
Sure, so, just for the listeners, I'm a cardiologist by
training and practice. I trained in the Midwest at the
University of Chicago, and then in two thousand and five
I took my first job as an attending cardiologist at
the New York Presbyterian Hospital and the wild Cornell Medical
College on the Upper East Side of New York City. There,
you know, we had my roles were twofold. I did
(01:26):
research and mostly with a focus on large scale research
and clinical trials, and then I also saw patients. The
focus of the research was really to try to help
us learn how to take better care of patients, and
so we had adopted a new technology that had just
come out in two thousand and five called the sixty
four slice CT scan. CT is just sort of a
(01:48):
fancy X ray, is what it is, but it, for
the first time gave us the ability to look at
certain diseases or heart disease in a different way than
we've never been able to do within the non evasive setting.
We didn't actually know what we were looking at at
the time, and so what we did was a series
of large scale clinical trials to see what we were
seeing on the image, how did that impact somebody's outcome,
(02:10):
and could we intervene to favorably influence the outcome so
that patients could be prevented from having heart attacks and
so on. We learned quite a bit over time, and
so we started a clinical program to infuse those research
lessons that we learned. We started that about twelve thirteen
years ago, and what we realized was that we wanted
(02:30):
to take a very different approach towards prevention and to
better precision medicine or personalized medicine. So we would leverage
the imaging to understand exactly how disease somebody was to
guide our treatment decision making, and then we would repeat
that every few years to make sure that we were
(02:50):
actually making good progress to eradicating somebody's risk of heart
attacks and other major adverse events. That was quite successful,
but quite manual, taking us maybe eight ten hours to
analyze a single CT scan, and so we realized though
with the clinical success that it was a much needed
tool to really try to personalize medicine and offer precision
(03:16):
care to help prevent adverse events from occurring. But we
had no way to scale it, and so we started
clearly to not really as a test. We didn't want
to make a single test. We didn't want to make
an AI algorithm. We do offer tests, and we do
offer all of our stuff is infused with AI, but
we more wanted to make a standardized care management platform
(03:38):
for every single stakeholder in the care pathway. So we
started clearly now a little over eight years ago, with
the intent to make a comprehensive care pathway for evaluation, education,
treatment and tracking for heart disease.
Speaker 1 (03:52):
Okay, And what it caught my attention there was just
how CT scans can make a personalized view for the patient.
But maybe why don't we walk through just the history
of angiograms in general, because you know your work over
the last twenty years and before your time, Yeah, how
(04:13):
have what have been the tools for physicians to help
diagnose cornary artery disease? And then with that said, as
the trend has gone over the last thirty forty years,
what are some of the setbacks that some of these
current diagnostic tools have.
Speaker 2 (04:30):
If you think about it, I think you can think
about sort of the life course of cornary artery disease.
What is corner artery disease. It's the plaque that builds
up in the heart arteries over time. It's the cause
of heart attacks, the number one cause of death. So
what we do and patients who are asymptomatic Historically we've
checked risk factors of heart disease. That's not heart disease,
(04:50):
but it's things like cholesterol and blood pressure and diabetes.
And we ask about certain social habits such as smoking
or things like that, and that's pretty much how we've
evaluated asymptomatic people. The issue with that is twofold One.
None of those things are heart disease, right, They're surrogate
markers of heart disease. And number two, they correlate very
(05:13):
unreliably with disease. So you can have patients with very
high cholesterol who have no disease, and conversely, you can
have patients with very low cholesterol who have very severe disease.
So it's not a great marker. And then what we
said was, well, you know, we usually send the patients
home and say, why don't you come back if you
have symptoms? What are those symptoms? Symptoms of chest pain
(05:35):
or shortness of breath and things like that. And historically
what we would do is we would perform stress testing
on these patients. And stress tests come in all forms
and flavors, but we would do our stress tests. Typically
the majority of them is what's called a nuclear stress test.
And somebody runs on a treadmill, We take some pictures
of their heart, we see what happens, what their heart
(05:57):
looks like at rest, and then what their heart looks
like when they're running, and we try to compare the two.
That's not heart disease either, Right, that's a downstream sequela
of heart disease, and then if that person has an
abnormal stress test, we would send them for what's called
an invasive angiogram. So the invasive angigram you can think
of as a shadow of the heart artery, right, So
(06:19):
if I took a flashlight too, like a computer monitor,
it would be what's behind the monitor the dark areas,
and what we're looking for there is whether or not
there's any narrowings in the heart artery. So you always
hear patients say, oh, my doctor told me I have
a blockage, But fundamentally, it's not the blockage that's the problem.
It's the plaque that causes the blockage. And we've never
(06:41):
been able to actually image that plaque non invasively at scale.
We could do that with invasive procedures, but we never
did it with non evasive procedures. So we went into
all of this research now dating back into two thousand
and five with the thought that we didn't actually understand
and heart disease because heart doctors have actually never measured
(07:03):
heart disease. We've measured upstream surroguate markers and downstream sequela,
but never the disease itself, So we just wanted to
understand the disease, and that's what the CT allows you
to do, is to not only look at the inside
of the artery where the blood is flowing, but also
the wall of the artery where the plaque is building up.
And what we found was that there's many, many different
(07:23):
kinds of plaque and they all have very different behaviors.
And if we can understand that, then we can understand
the patient better and their risk of having a heart attack.
Speaker 1 (07:33):
Okay, and so you know, when you have a typical
heart patient come in, what's the steps that gets them
to eventually to a CT scan? Because I don't I'm
assuming that if I go to my primary care physician,
they're not going to waste their time on me with
a CT scan. How far in the process does it
go until they get to a CT scan? And do
(07:55):
they have to do like other scans too, whether it's
a traditional angiograph, free scan, angeogram, think anything like that,
and then they go to the CT scan.
Speaker 2 (08:06):
Good question, Like I think, you know, when you look
at the other technologies that we use in cardiology, the
two that you mentioned, a stress test or an angiogram,
those have been around for a long time, right, so
they've been around. I think nuclear stress testing has been
around for about fifty years. Invasive angiography has been around
for longer than that. So the CT is relatively new.
(08:28):
I mean, it's still twenty years old, but it's relatively
new compared to the older tests there was at the
time it was introduced. Like, we as a field of
cardiology pride ourselves on evidence based medicine, so things really
need to be proven out in large scale clinical trials.
There's been now since it was introduced a few dozen
(08:49):
large scale randomized control trials with CT, all which have
proven its worth and its superiority over other methods. So
in twenty twenty one, the American Heartssation and the American
College of Cardiology released the Chest Pain Guidelines or a
guideline Professional guidance statement that stated that CT and gigraphy
(09:11):
is Level one A. That's the highest recommendation that it
can get, and it was the only recommendation for patients
who are symptomatic was suspected corner artery disease to receive
that non designation. So if you had a feeling of
symptoms in your doctor, suspected corner artery disease. Likely you
might get a prescription for a CT aigram or a
(09:33):
sixty four slice CT scan or a CCTA whichever. These
are all the same words or acronyms and abbreviations for
the same thing, which is a totally non invasive tool
like so that means that you could go as a
compared to the invasive androgram, which intrinsic in its name.
It's invasive and you have to put some needles in
(09:53):
and thread catheters up to the herd arteries and things
like that. The non invasive CT androgram is very simple. Again,
a CT is just a fancy X ray in a
very high quality, highly efficient site. You'd probably be in
and out in fifteen twenty minutes or so, completely painless.
So it's quite easy to perform, and it's quite easy
(10:15):
for the patient to undergo. And I think it just
gives you the richest amount of informations as endorsed by
the professional societal guidelines.
Speaker 1 (10:24):
Gotcha, and then kind of just circling back to your
introduction maybe and I just to connect all the dots together.
The one of the big barriers though, is that the
time it takes us to review and analyze a single
scan correct. And when you're talking about it, you know,
sometimes it's eight to ten hours to review a single scan.
(10:46):
That's kind of one of the barriers with the traditional
CT scans, right.
Speaker 2 (10:52):
I think that the way that we typically did it
was that somebody would sort of eyeball it, right, so
very subjective, very non precise. And what we find is
that readers, including myself, tend to always overestimate things and
then overestimate the severity of blockages and then completely misclassify
(11:16):
the plaques that are building up in the walls of
the artery. So we recognize that the need. The initial
step was to make things more efficient so you could
actually scale it and deliver to patients, and to improve
the accuracy so that we could actually provide real, actual
numbers to patients that were meaningful from their own bodies.
(11:37):
That's where the AI helped. So we had been working
on this for quite some time, and then in twenty ten,
there was this thing called machine learning, and we didn't
know what it was at the time, but we said,
you know, it sounds new, so why don't we at
least sort of suss it out and see whether or
not it works. It didn't work very well at the time,
It works really well right now. So it turned out
(11:59):
to be a good decision that we allocated resources because
we got in sort of at that early stage during
the evolution of AI and machine learning. Yeah.
Speaker 1 (12:07):
No, that's interesting. And before actually, before we jump into that,
one of the things I heard from you is it
sounds like there's you the doctors would tend to be
on the safer side, so there would be tend to
be more false positives when they're doing their kind of
subjective eyeball analysis, right.
Speaker 2 (12:22):
Yes, and that's been proven time and time again. Like
we've seen in virtually everything, people tend to air on
the side of conservatism. So what is that I'd rather
call something positive and be wrong than call something negative
and have a catastrophic event happen to the patient.
Speaker 1 (12:39):
Yeah, no, that's it. Yeah, And that's I think that's
why having something that's more objective is very important for
these type of scans that's also done in a timely manner.
Speaker 2 (12:48):
The problem with the false positives that we have for
these kinds of tests, whether it's stress tests or cctas
or whatnot, is that you can air on the side
of conservatism. But by calling it a false positive, what
you do is you end up sending patients to the
next downstream test, which happens to be a very invasive,
very expensive procedure. So that just caused a lot of
(13:11):
unnecessary harm to the patient and a lot of unnecessary
waste to the healthcare system. So we're trying to prevent
all of that.
Speaker 1 (13:19):
Yeah, you rather well measure twice and cut ones type
of c scenario actly. But that kind of gets to
that machine learning because you're you're you, you nailed the
hammer on the nail there where the machine I've heard
machine learning, I feel like since I first started in
the early Yeah you almost ten plus years ago now,
(13:41):
but it never really caught wind until the rise of AI.
And then when you talked about how you started clearly
eight years ago, that was half a decade before all this,
you know, the huge AI bump that we were seeing
over the last two three years. So what were your
initial feedback of that machine learning and kind of what
(14:04):
were the challenges and what were the steps that you
needed to do to kind of overcome those you know,
those challenges.
Speaker 2 (14:11):
The first thing we had to do was learn it,
and so that was pretty tough. Like, we had a
technical computational biology group where I was and you know,
it probably took us a year and a half to
just speak the same language. Right when I would say
something clinical or they would say something technical. We sort
of missed each other for about the first year, year
(14:32):
and a half. Then somehow it clicked where we could
understand they understood the clinical problem to be solved, I
understood the technical approach that they were taking, and so on.
I mean, we've seen iterations of this. I think when
we started the company, that was when the whole sort
of deep learning convolutional neural nets were there that was
good enough, right, Like that was a huge step forward
(14:53):
in terms of image processing to higher accuracy and efficiency.
So we sort of luck doubt at the time because
if we were going with traditional image processing probably wouldn't
the solution wouldn't be half as good as it is today.
And now things have improved and they continue to improve. Also,
what also improves is like the corpus of data that
(15:15):
we can train these algorithms on and then validate them.
You know, we've increased those over time, so we've got
hundreds of thousands of cases that we can access to
really train models in a way that they can recognize
any type of heart disease and really try to offer
this kind of personalized approach to preventing heart attacks. So
we try to support the doctors that way.
Speaker 1 (15:36):
Yeah, and then maybe let's just dive a little deeper
into clearly itself, what the product is, what it does,
how the algorithms work. And then what's interesting is that
you've expanded the offering and so there's now and then
if I understand it correctly, there's clearly, but there's also
clearly a schemia, and there's clearly plaque analysis, and it's
(15:56):
just kind of expanded out. So how is the whole
repertul are of products work, either alone or together to
help you with the patient scans?
Speaker 2 (16:08):
Yeah, it's a good question. Like it goes back to
sort of what we talked about earlier. Again, we wanted
to just make a comprehensive care management platform. What does
that mean? That means that in a single user experience,
I as the user, can get everything that I want
and everything that I need to best take care of
that patient. So who is the definition of I. Well,
(16:29):
there's many eyes, right, there's the imager or the radiologists
who has to read the images. There's the general cardiologists
who've may be taking care of the patient. There's the
preventive cardiologists who more focused on sort of personalizing assessment
and treatment of anthroscrosis. There's the interventional cardiologists who takes
people to the invasive lab and puts in stents. There's
(16:51):
the researchers, there's the primary care physicians and the advanced
practice providers, and most importantly, there's the patient. So the
way that we construc DIDUD this was the first thing
that we made was the care management platform. So if
we're a software as a service company, you just log
onto any onto the Internet from any computer and you
can access this platform that gives you exactly the data
(17:14):
that you need as the stakeholder. So the interventional cardiologists
needs something different than the radiologists. The radiologists need something
different than the primary care doctor, and so on and
so on. So this is the platform that was the
overall structure, and it comes in a way that translates
all of this fancy imaging stuff and pixels into stuff
that doctors can actually understand and patients can actually understand.
(17:37):
So that what we felt was a very important part
of the platform, the translation step, because when we were
at Cornell and New York Presbyterian, we would talk to
all of our colleagues about all this fancy imaging findings
and they're like, I don't know what you're talking about,
because they didn't They shouldn't, right, They weren't imagers. They
were clinicians in the office taking care of patients. So
we realized that all of this important data was getting
(18:00):
lost in the handoff, and so we wanted to rescue
that data by creating this platform. And then after that
we have three additional so we have four fully FDA
cleared products. The first one that one is the care
management platform and the experience, and then the other three
ones are The first one is called Clearly Labs. We
call it Clearly Labs. And what that looks at is
(18:22):
the plaque that is building up very comprehensively in every artery,
in every vessel, in every segment and every lesion. We
can characterize quantitatively all of the differences of the plaque
that's silently building up in the walls of the artery
and putting the patient at risk of a heart attack.
We can also calculate the narrowings in the arteries that
are caused by that plaque. That's called stenosis or blockage
(18:46):
or narrowing. And then we have a second product that
takes all of that information and leverages end to end
artificial intelligence to determine the hemodynamic significance of those lesions.
What do I mean by humanemics significance is those things
cause a reduction in blood flow when say somebody's exercising
or walking up the stairs. That may be the cause
(19:07):
of somebody's chest pain or shortness of breath. So then
what you get are three things. You get this athriskrosis
or plaque that's building up in the wall. You get
the stenosis or the narrowing caused by the plaque, and
then you get whether or not that stenosis is actually
impeding flow on blood flow in a way that may
cause somebody's symptoms. So those are the three major categories.
(19:30):
Within those categories are a ton of subcategories, but the
three major categories of actionable findings for cornerary artery disease
where you know how to act upon it to try
to improve a patient's outcome. And then the last product
that we have really fits into that care management pathway
that we wanted to do around evaluation, education, treatment, and tracking.
(19:52):
It's a quantitative tracking software that we call clearly Compare,
So in the case that somebody had two scans, we
can tell you what's changed in that patient's heart over time.
That's very important because when we give somebody a medication
like a statin, for example, to lower their cholesterol, what
we see is a twenty percent reduction in heart attack events.
(20:15):
That sounds great, except that means that you've lowered cholesterol
in one hundred people and eighty percent of them aren't
benefiting from the drug. We call that population the residual
risk population. But I don't know who they are. I've
given them the drug, the cholesterol is lower, they're seemingly better,
and yet four out of five people aren't benefiting from that.
(20:36):
We think that quantifying disease changes over time will be
could be a very effective way to figure out who
those people are who are at residual risk. And then
it's just a good way to see whether or not
the medicines are working or not, or whether or not
disease is still progressing.
Speaker 1 (20:52):
No, it makes sense because you want to beform an
after and you want to see how the intervention is
actually changed or help the patient's heart. Let me just
ask because it just came to my mind. So you
have all these different products. Are they all packaged into
one commercialize product that you sell to the hospital or
(21:14):
to the radiologists or is there a kind of ALEC
heart menu for these customers to pick and choose which
ones work best for them and their and their hospital budget.
Speaker 2 (21:26):
It's a great question. I mean, it depends on the client.
We you know, I favor sort of all for one
kind of thing, Like I think that comprehensive care is
better than fragmented care, standardized care is better than non
standardized care. So if it were me, I want everything
I want to know about my heart. Some sites adopt
(21:48):
that approach, and we do that through a subscription model.
Others do not, and they would prefer to go all
a cart. We'll meet the client wherever their needs are.
The one thing you know, as you well know, is
you know, for the site to get paid, they typically
have to go through some insurance coverage and reimbursement. And
(22:08):
so we have two codes right now that they're called
CPT codes or current Procedural Technology codes. That's how the
site can build the insurance company so that the patient
doesn't have to pay out of pocket. And so sometimes
depending on how they build and how they submit insurance
claims and so on, sometimes the all the cards better
(22:28):
for them. Sometimes the all for one is better for them.
We just try to work with them to maximize delivery
to as many at risk patients as we can.
Speaker 1 (22:37):
Yeah, and you know, one of the things, you know,
we'll get into reimbursement a little bit more because I
know it's a very evolving landscape in the CMS space.
But I'm curious as well as you know, you talk
about reaching as many patients as possible. There's a plethora
of clinical trials out there teaching, you know, studying different
(22:58):
patient cohorts. And I know one of the most recent
ones is confirmed too, so we can start with that one.
But what are some of the clinical trials clinicals results
when you have the product when it first came out,
that are kind of the important Yeah, study points for
(23:19):
this technology.
Speaker 2 (23:21):
Yeah, it's a great question. There was a we sort
of follow a roadmap. That roadmap is like does this
product work? That's the first question we ask, And so
that's usually a diagnostic performance or diagnostic accuracy question. So
what we've done is we've tested our tools against every
quote unquote gold standard that we can possibly think of.
(23:44):
That turns out to be the invasive methods of imaging,
which have you know, they're closer to the hard arteries
because they're inside the arteries and so they have very
high resolution. So we've tested against five different invasive modalities.
I won't go in to all of the names of them,
but we do that with blinded core laboratories against these
(24:05):
gold standards and have demonstrated high performance. The second question
that we ask is is what I'm detecting even relevant
doesn't matter and there are tests out there that if
they're normal or they're abnormal, it doesn't really affect your life.
So then the question is why are you measuring it?
(24:25):
Because if it's not prognostically useful, then it's not a
particularly useful test. And so we've done a series of
large scale studies with two, three, four, five, seven, eight,
and ten year outcomes to demonstrate that when we see something,
we can identify people at risk in a very graded
fashion so that we understand who needs to be treated
(24:48):
more aggressively. The third thing that we've done is a
series of large scale clinical studies that have eight, seven
and eight year follow up to make sure that when
people are acting upon our study findings that the patient's
outcome can actually be improved, that we can reduce future
major adverse cardiovascar events, things like heart attacks, death stroke,
(25:09):
and things like that. So we've just presented two large
scale studies at the American College of Cardiology meeting a
few months ago. The fourth thing that we do is
to try, you know, we should be stewards good stewards
for the healthcare system is extremely big spend for the
US government and for patient families that are paying commercial insurance,
(25:32):
and so we want to make sure that our tools
don't just add bloat to the healthcare system, but actually
have the potential to actually significantly bend the cost curve downwards.
So we've spent a significant amount of time publishing some
studies that demonstrate cost savings associated with our tools as well,
and so those that's sort of the tiered approach that
(25:55):
we've taken to try to prove our stuff. Coming back
to what you asked about, we believe like all of
the stuff that went into clearly was based on science, right,
all these large scale clinical trials. So that's our DNA, right.
We ran clinical trials for twenty years, and so we
are committed to continuing that science evolution and sort of
(26:21):
searching for always the right answers, and we will continue
to learn for the rest of our lives, because that's
just the iterative process of complex human biology. But the
easiest way you can learn that is from very high quality,
large scale studies. And so the one that you suggested
(26:41):
was Confirmed two. The reason it's called Confirm two is
because when I was at Cornell, we did Confirm one.
That was a study of twenty three thousand patients who
we had followed for two years. Confirmed two goes it's
like Confirm one on steroids. So what we have done
is now we've amassed fifty some sites across the globe
(27:05):
from multiple different ethnicities and countries, and what we have
done is about five year follow up of these patients
who have undergone everything that clearly has to offer, so
advanced plaque analysis, advanced narrowing analysis, advance the schemia analysis,
and we've adjudicated all of the outcomes, and so it's
(27:26):
just teaching us a lot about not only clearly tests
and how that impacts somebody's outcome, but also just Vasker
biology in general. We're learning brand new biological concepts. As
one of them we presented at the late breaking clinical
trials for the American College of Cardology was gender differences.
We fully don't fully understand the difference between women and men,
(27:49):
but when you look at the data, Wow, women and
men behave extremely differently when it comes to heart disease,
and we need to recognize that in order to better
take care of women who we've sort of lumped into
the same category as men for too many years. And
so it's just that one is we'll advance our understanding
of science, and then it's really going to validate the performance,
(28:12):
the prognostic performance of our products in many, many studies
to come.
Speaker 1 (28:18):
Yeah, and that's the thing about confirmed too that caught
my attention is it was a fifty to fifty split
between men and women, when usually these clinical trials are
have the women are well underrepresented based on you know,
relative to kind of the actual treatment options that are available.
What caught my attention to as well. And actually we'll
probably cover that with the reimbursement side, but The other
(28:43):
one too that's coming out that's currently enrolling, is the
Paramount trial. And what caught my attention is that's a
randomized trial. So what's the importance of doing that to
help with you know, making you know, CT scans and
the AI funk features with these CT scans become more
(29:04):
prevalent at these hospitals.
Speaker 2 (29:05):
Yeah, it's a great question. So we have two large
scale randomized control trials ongoing. One of them is Paramount,
the other one is called Transform, and I'll tell you
the rationale behind both of them. The you know, for
the listeners, like, a randomized trial is where you take
a population of people and you subject in general, half
(29:26):
to one thing and half to another. And that's to
just alter one variable in the experiment and then to
see what happens and what changes in the outcome. That's
the sinaquonon for cardiovascar medicine. Every you know, we cardiology
is the most evidence based field and it has relied
on the mechanism of the randomized trial. And it's because
(29:49):
it's a very controlled That's why it's called a randomized
controlled trial. You're just changing one variable and you see
what the outcome is. So what's the difference between that
and say, like a confirmed to confirm two is an
observational cohort study that means you just take ten thousand
consecutive people and you watch them, and that's called a registry.
So the randomized trial is great for proving one thing,
(30:13):
but there's always inclusion criteria and exclusion criteria, so it's
not everybody. So to me, what randomized controlled trials does
is proves your point and really gets you guideline quality
data to change the American Heart Association American Culture Cardiology guidelines.
And what registries do is it teaches you how to
use it in clinical practice, because those are all comers,
(30:36):
and so we try to do both of those. So
for the two randomized trials that we're doing, one of
them is called Paramount. It is a randomized trial of
patients who have symptoms suggestive of heart disease, chest pain,
shortness of breath, and so on. And what we wanted
to do was really look at the influence of what
clear these technologies could do to improve patient health outcomes,
(30:58):
to reduce seri invasive procedures, to improve medical or cornority
disease resk fact or control, to really show that like
this is good for patients, it's good for the healthcare
system at large. And so yeah, that trial, we're aiming
to finish enrollment by the end of this calendar year,
and that will hopefully prove the role of end to
(31:21):
end AI within a single, unified, comprehensive care management platform
to improve patient health outcomes. The second trial is called Transform.
It's a much bigger swing at the bat. So if
you look and you asked me, what is the biggest
unmet need in coronary heart disease the cause of heart attacks,
(31:41):
I will tell you it is the asymptomatic people walking
on the street. And you'll scratch your head and say,
I don't understand that. Well, it turns out everybody thinks
of heart attacks as having chest pain and so on.
That's very common, and everybody thinks that the precursor to
having a heart attack is having chest pain for a while,
and then you don't you know. But what happens when
you look at the data is that somewhere between fifty
(32:04):
and sixty percent of people who will die from a
heart attack or suffer a heart attack never have a
symptom before their event. And I think you and I
know everybody knows someone like that, right who went for
a run and died and never came back, or went
to sleep and never woke up. And the reason everybody
knows something like that is that it's the number one
(32:24):
cause of death in this world. And I can be
the best preventive cardiologist in the world in you know,
at New York Presbyterian Hospital or wherever, but those patients
don't come to me because they die in the home.
So the biggest unmet need is to find those asymptomatic
people who comprise the majority of people who will have
heart attacks and go find them. And I think the
(32:45):
only way you can do it is to go up
into the home. And so we have this trial called TRANSFORM,
a randomized control trial for screening completely asymptomatic individuals. They're
not patients, they're not people in the hospital. They're the
people that you see walking in the grocery restore and
playing Little League on the weekends who then die suddenly
of heart attacks. And so what we're it's a seventy
(33:07):
five hundred patient study. We're about thirty seven thirty eight
hundred patients enrolled, and we are testing the hypothesis that
imaging actual disease of a patient is better than checking
risk factors of disease like cholesterol. So you know, we're
we're still we're enrolling as quickly as we can because
(33:29):
I think this is the number one public health epidemic.
Like in twenty twenty, there are twice as many cardibaskard
desks that than desks from COVID nineteen. That's how big
this epidemic is. And so we feel very motivated that
there's an urgency to us to find the answer to
better take care of patients.
Speaker 1 (33:44):
There's a lot to unwrapped there, because you're totally right,
and you thought that, you know, I would be surprised
when I hear about you know, fifty sixty percent, you know,
don't have any symptoms before. But you hear countless stories
either at a personal level or you know, just in
the news and things like that, and then all you
just hear is like, oh, they just dropped dead. They
were perfectly fine, you know, as you said, they were running,
(34:05):
and then just it's unfortunately, you're right, it's so it's
so tragic about how instant that could be.
Speaker 2 (34:12):
Just a short story, Like two weeks ago, I was
talking with a doctor and she had ordered a number
of clearly analyzes on her patients, and she was just
showing me, like what she one patient, Like she had
deidentified the patient, but she said that he was pretty young,
he was in his mid fifties, and just showed me
(34:33):
the clearly analysis findings and they were pretty marked, and
I thought, wow, this person is quite sick. Like what
did you do with this patient? And she said, well,
I recommended that he go on certain medical therapy and
I said, yeah, I think that. I mean, just as
a former practicing cardiologist, that sounds pretty reasonable. And I
said what happened? She said, he refused it. He said
(34:54):
he didn't want to take any medications. So they agreed
that they would talk about it the following week and
just to see what the plan should be. He died
on Monday, like four days later. So it's like those
kinds of things that just make you so humble that like,
we really need to tackle this problem. It's and that
person was so young in their mid fifties, Like there's
(35:15):
just too much of that. So that was our north
star when we started the company, that the tool we
wanted to make was a comprehensive care management platform, But
our north star was really to try to accrue the
evidence and prove or disprove the utility of just measuring
disease over risk factors of disease to prevent heart attacks.
Speaker 1 (35:35):
Yeah, and let's just let's dive into that because I
know in the past you've mentioned that you would want
to imagine a world where there's no sun in death
from heart attacks, and here we are just talking about
those instant, tragic events. But there's a lot of steps
that need to get there, and well, maybe you can
just jump into each of those steps. But one of
(35:56):
the first ones that come to mind is the cost
of a CT scan versus some of the other traditional
methods of scanning for patients. And there's also the reimbursement landscape.
So walk us through. You know, the cost difference is
the reimbursement because, like you said, sometimes the hospital gets
a certain reimbursement and they only can only do so
(36:18):
many scans, they only do one or two of your
comprehensive skins. And then talk about the you know, costs,
economics and the data that you're seeing that could say, hey,
it might be a little bit more now, but in
the long run, this is the right decision to make.
Speaker 2 (36:37):
Yeah, it's both of them are really great questions. So
let me be as clear as I can. We are
not advocating right now for universal screening at all. And
the coverage and reimbursement that we have for our tools
is solely restricted for symptomatic patients with suspected heart disease,
shortness of bread, chest pain, and so on, and that
(36:59):
is currently where our company is focusing its commercial efforts
on that symptomatic population. So I don't want it. I
want it to be very clear that we are not
here advocating for universal screening until or unless this large scale,
randomized trial is positive, and at that point in time,
then we'll start to lobby all of the insurance providers
(37:19):
to say, look, we can save lives here. So for
the first one, that's the symptomatic population, I think that's
a traditional fee for service model. What's that mean? Like
I do a service, you give me a fee, the
insurance company gives me if the hospital fee and so on.
There's also value based models that we I think fit
very well in where we can be a better personalized
(37:42):
referral management tool to reduce costs on the screening side.
Which is I think where you were directing some of this.
You know, there's a number of ways that you can
approach this. They're all long, they're all very expensive, and
they all require level grade A science. And if you
don't have that, then I don't think you have a
prayer to try to move this kind of movement forward,
(38:05):
particularly as big a frameshipt as we're trying to espouse.
When you look at the data though, and you look
at the estimates, when we said, okay, what are we
going to see in our clinical trial, you have to
come up with a hypothesis we're going to reduce heart
attacks by ten percent, twenty percent, thirty percent. And so
when you look at that, we talk about a number
(38:25):
needed to screen in order to identify and prevent a
bad thing from happening. For mammograms, the data that's been
published is that you need about six hundred women to
undergo fifteen consecutive years of mammograms in order to prevent
one breast cancer death. For colonoscopies, it's around thirteen hundred
(38:46):
and fifty patients if our number needed, If our hypothesis
is correct for this trial, the number needed to screen
would be thirty so if you believe that screening mammograms
and screening colonoscopy are cost effective and that we should
be doing them, then this should be wildly cost effective
(39:07):
if the trial is actually positive, right, and then in
terms of the price, fortunately or unfortunately, the cordinerary CT
indrogram is not particularly expensive, Like it's one of the
lower cost advanced imaging tests out there. So all in,
if you were to bake in the CT scan costs
(39:28):
and the clearly analyzes certainly be less than a colonoscopy
which we're using for universal screening, and then it would
identify a lot more people at risk. So you know,
we have to finish the trial. We have built in
very complex advanced cost effectiveness analysis within the trial, so
we'll we'll provide all of that data as well. But
(39:50):
there's just too many people dying of heart attacks, Like
we've got to do something. And when you said, well,
where's the problem, I see three different legs of this
stool identification, treatment and implementation. If you look at the
treatment like just currently what we have today in our toolbox,
which is about a dozen different medications and lifestyle modifications.
(40:12):
If you do the additive relative risk reductions over ninety percent,
we can eradicate heart attacks. What we don't do is
identify well. I think that's where it clearly really comes
in to hone in on a precision analysis of somebody's heart,
and then we don't implement well because we wait for
people to come to the hospital in the throes of
some catastrophic event rather than to try to prevent it
(40:34):
early when it can be much more easily prevented. So
this whole like we're trying to address those two legs
of the stool, the identification and the implementation.
Speaker 1 (40:44):
Yeah and yeah, it seems like you're in your adding
the to RCT, using especially transform is definitely the right
step because you're right, sometimes when you wait for the symptoms,
it's too late at that point. So it sounds like though,
I mean, right now you're advocating for only symptomatic because
you're waiting until the RCT results come out and it
(41:06):
hopefully supports your hypothesis to start training the sooner if
everything goes as you and vision, How do you see
this technology being utilized over the next five to ten years,
And especially what caught my attention to is you talked
about iterative not only for the data for the AI
tools to keep learning, but also the human aspect of it.
(41:29):
As you were saying, you're always going to continue to learn.
How do combining both of those segments together create what
AI truly is designed to help society out with?
Speaker 2 (41:43):
I mean, like going forward. I can answer that in
a number of different ways. Like, we have a current
set product roadmap that's over ten years long, so we're
nowhere near done. It will likely change over time because
we'll be influenced by new and inovations that come about.
And what's embedded what's not embedded in that product roadmap
(42:07):
is sort of the power of what AI can offer,
which is inferences from data and data at scale. Right,
So you know, I believe for heart disease that seeing
heart disease, actual heart disease is the most effective way
to identify somebody who is sick and to guide your treatment.
I believe that in my heart of hearts. That doesn't
(42:30):
mean that other data is not important. Right. So there's
plenty of particularly in the cancer space. There's these liquid
biopsy companies, right, and some look at genomics and polygenic
risk scores, some look at epigenetics, some look at proteomics,
some try to combine them altogether and so on and
so on. Or there are companies out there who are
focusing on biometrics that you get from wearables, your rr
(42:52):
A ring or your Apple watch or your garment. I
think they're all useful, I think to a lesser degree
than whether you actually have heart disease. But I think
the future of AI is that we will integrate all
of these seemingly disparate data and we will output it
in a way that would be clinically actionable. To that
(43:12):
you just need a ton of data, right, Like that
means you need blood samples and you know what your
watch said and did you sleep well, and what's your whatever.
So like that's going to be take some time to
approve that kind of really high quality data, and then
it's also going to require a very respectful approach to
(43:33):
regulatory approval and to coverage and reimbursement, right, So those
are things, and then it's got to be proven a
science and proven through large scale clinical trials. So like,
I don't ever worry that AI is going to outpace us,
because I think AI is just a tool. It's a
(43:55):
tool that's embedded into a medical device, and because it's
a medical device. It requires the same careful, meticulous approach
towards regulatory approval, towards evidence generation, towards large scale clinical trials.
And so I don't think the field of medicine ever
will or should outpace the AI technology improvements because the
(44:17):
stakes are too high. Right, the AI gets it wrong,
somebody dies, that's just not acceptable. So you have to
just keep testing it in an evidence based fashion to
prove beyond certainty that the device works and it actually
improves patient healthcare.
Speaker 1 (44:33):
Yeah, and you know what, at the end of the
day that all works out. That's how you optimize AI
in this world. But Jim, thank you so much for
joining us today. It was a very enlightening episode.
Speaker 2 (44:43):
No, it's a pleasure to be here, and thanks so
much for having me on the show.
Speaker 1 (44:47):
Yes, and thank you to our listeners for tuning in today.
We hope you join us for future episodes, and if
you'd like to stay up to date, you can click
the subscribe button on Spotify or your favorite streaming platform.
Take care Chasser uses US
Speaker 2 (45:24):
Changes bases
Welcome to another episode of the Vanguards of Healthcare series.
My name is Matt Hendrickson, the medical technology analyst at
Bloomberg Intelligence, which is the in house equity research platform
of Bloomberg Elpeep. We're pleased to have with us today
doctor Jim Mint, founder and CEO of Clearly, a privately
held medical device company that is applying artificial intelligence to
(00:40):
non invasively measure plaque stenosis in the likelihood of a
schemia within the heart. Jim, thank you for joining us today.
Speaker 2 (00:49):
Thank you for having us man like, it's a pleasure
to be here.
Speaker 1 (00:52):
Likewise, and why don't we just start with an overview
of your career path and then the steps in that
path that led you to founding Clearly.
Speaker 2 (01:02):
Sure, so, just for the listeners, I'm a cardiologist by
training and practice. I trained in the Midwest at the
University of Chicago, and then in two thousand and five
I took my first job as an attending cardiologist at
the New York Presbyterian Hospital and the wild Cornell Medical
College on the Upper East Side of New York City. There,
you know, we had my roles were twofold. I did
(01:26):
research and mostly with a focus on large scale research
and clinical trials, and then I also saw patients. The
focus of the research was really to try to help
us learn how to take better care of patients, and
so we had adopted a new technology that had just
come out in two thousand and five called the sixty
four slice CT scan. CT is just sort of a
(01:48):
fancy X ray, is what it is, but it, for
the first time gave us the ability to look at
certain diseases or heart disease in a different way than
we've never been able to do within the non evasive setting.
We didn't actually know what we were looking at at
the time, and so what we did was a series
of large scale clinical trials to see what we were
seeing on the image, how did that impact somebody's outcome,
(02:10):
and could we intervene to favorably influence the outcome so
that patients could be prevented from having heart attacks and
so on. We learned quite a bit over time, and
so we started a clinical program to infuse those research
lessons that we learned. We started that about twelve thirteen
years ago, and what we realized was that we wanted
(02:30):
to take a very different approach towards prevention and to
better precision medicine or personalized medicine. So we would leverage
the imaging to understand exactly how disease somebody was to
guide our treatment decision making, and then we would repeat
that every few years to make sure that we were
(02:50):
actually making good progress to eradicating somebody's risk of heart
attacks and other major adverse events. That was quite successful,
but quite manual, taking us maybe eight ten hours to
analyze a single CT scan, and so we realized though
with the clinical success that it was a much needed
tool to really try to personalize medicine and offer precision
(03:16):
care to help prevent adverse events from occurring. But we
had no way to scale it, and so we started
clearly to not really as a test. We didn't want
to make a single test. We didn't want to make
an AI algorithm. We do offer tests, and we do
offer all of our stuff is infused with AI, but
we more wanted to make a standardized care management platform
(03:38):
for every single stakeholder in the care pathway. So we
started clearly now a little over eight years ago, with
the intent to make a comprehensive care pathway for evaluation, education,
treatment and tracking for heart disease.
Speaker 1 (03:52):
Okay, And what it caught my attention there was just
how CT scans can make a personalized view for the patient.
But maybe why don't we walk through just the history
of angiograms in general, because you know your work over
the last twenty years and before your time, Yeah, how
(04:13):
have what have been the tools for physicians to help
diagnose cornary artery disease? And then with that said, as
the trend has gone over the last thirty forty years,
what are some of the setbacks that some of these
current diagnostic tools have.
Speaker 2 (04:30):
If you think about it, I think you can think
about sort of the life course of cornary artery disease.
What is corner artery disease. It's the plaque that builds
up in the heart arteries over time. It's the cause
of heart attacks, the number one cause of death. So
what we do and patients who are asymptomatic Historically we've
checked risk factors of heart disease. That's not heart disease,
(04:50):
but it's things like cholesterol and blood pressure and diabetes.
And we ask about certain social habits such as smoking
or things like that, and that's pretty much how we've
evaluated asymptomatic people. The issue with that is twofold One.
None of those things are heart disease, right, They're surrogate
markers of heart disease. And number two, they correlate very
(05:13):
unreliably with disease. So you can have patients with very
high cholesterol who have no disease, and conversely, you can
have patients with very low cholesterol who have very severe disease.
So it's not a great marker. And then what we
said was, well, you know, we usually send the patients
home and say, why don't you come back if you
have symptoms? What are those symptoms? Symptoms of chest pain
(05:35):
or shortness of breath and things like that. And historically
what we would do is we would perform stress testing
on these patients. And stress tests come in all forms
and flavors, but we would do our stress tests. Typically
the majority of them is what's called a nuclear stress test.
And somebody runs on a treadmill, We take some pictures
of their heart, we see what happens, what their heart
(05:57):
looks like at rest, and then what their heart looks
like when they're running, and we try to compare the two.
That's not heart disease either, Right, that's a downstream sequela
of heart disease, and then if that person has an
abnormal stress test, we would send them for what's called
an invasive angiogram. So the invasive angigram you can think
of as a shadow of the heart artery, right, So
(06:19):
if I took a flashlight too, like a computer monitor,
it would be what's behind the monitor the dark areas,
and what we're looking for there is whether or not
there's any narrowings in the heart artery. So you always
hear patients say, oh, my doctor told me I have
a blockage, But fundamentally, it's not the blockage that's the problem.
It's the plaque that causes the blockage. And we've never
(06:41):
been able to actually image that plaque non invasively at scale.
We could do that with invasive procedures, but we never
did it with non evasive procedures. So we went into
all of this research now dating back into two thousand
and five with the thought that we didn't actually understand
and heart disease because heart doctors have actually never measured
(07:03):
heart disease. We've measured upstream surroguate markers and downstream sequela,
but never the disease itself, So we just wanted to
understand the disease, and that's what the CT allows you
to do, is to not only look at the inside
of the artery where the blood is flowing, but also
the wall of the artery where the plaque is building up.
And what we found was that there's many, many different
(07:23):
kinds of plaque and they all have very different behaviors.
And if we can understand that, then we can understand
the patient better and their risk of having a heart attack.
Speaker 1 (07:33):
Okay, and so you know, when you have a typical
heart patient come in, what's the steps that gets them
to eventually to a CT scan? Because I don't I'm
assuming that if I go to my primary care physician,
they're not going to waste their time on me with
a CT scan. How far in the process does it
go until they get to a CT scan? And do
(07:55):
they have to do like other scans too, whether it's
a traditional angiograph, free scan, angeogram, think anything like that,
and then they go to the CT scan.
Speaker 2 (08:06):
Good question, Like I think, you know, when you look
at the other technologies that we use in cardiology, the
two that you mentioned, a stress test or an angiogram,
those have been around for a long time, right, so
they've been around. I think nuclear stress testing has been
around for about fifty years. Invasive angiography has been around
for longer than that. So the CT is relatively new.
(08:28):
I mean, it's still twenty years old, but it's relatively
new compared to the older tests there was at the
time it was introduced. Like, we as a field of
cardiology pride ourselves on evidence based medicine, so things really
need to be proven out in large scale clinical trials.
There's been now since it was introduced a few dozen
(08:49):
large scale randomized control trials with CT, all which have
proven its worth and its superiority over other methods. So
in twenty twenty one, the American Heartssation and the American
College of Cardiology released the Chest Pain Guidelines or a
guideline Professional guidance statement that stated that CT and gigraphy
(09:11):
is Level one A. That's the highest recommendation that it
can get, and it was the only recommendation for patients
who are symptomatic was suspected corner artery disease to receive
that non designation. So if you had a feeling of
symptoms in your doctor, suspected corner artery disease. Likely you
might get a prescription for a CT aigram or a
(09:33):
sixty four slice CT scan or a CCTA whichever. These
are all the same words or acronyms and abbreviations for
the same thing, which is a totally non invasive tool
like so that means that you could go as a
compared to the invasive androgram, which intrinsic in its name.
It's invasive and you have to put some needles in
(09:53):
and thread catheters up to the herd arteries and things
like that. The non invasive CT androgram is very simple. Again,
a CT is just a fancy X ray in a
very high quality, highly efficient site. You'd probably be in
and out in fifteen twenty minutes or so, completely painless.
So it's quite easy to perform, and it's quite easy
(10:15):
for the patient to undergo. And I think it just
gives you the richest amount of informations as endorsed by
the professional societal guidelines.
Speaker 1 (10:24):
Gotcha, and then kind of just circling back to your
introduction maybe and I just to connect all the dots together.
The one of the big barriers though, is that the
time it takes us to review and analyze a single
scan correct. And when you're talking about it, you know,
sometimes it's eight to ten hours to review a single scan.
(10:46):
That's kind of one of the barriers with the traditional
CT scans, right.
Speaker 2 (10:52):
I think that the way that we typically did it
was that somebody would sort of eyeball it, right, so
very subjective, very non precise. And what we find is
that readers, including myself, tend to always overestimate things and
then overestimate the severity of blockages and then completely misclassify
(11:16):
the plaques that are building up in the walls of
the artery. So we recognize that the need. The initial
step was to make things more efficient so you could
actually scale it and deliver to patients, and to improve
the accuracy so that we could actually provide real, actual
numbers to patients that were meaningful from their own bodies.
(11:37):
That's where the AI helped. So we had been working
on this for quite some time, and then in twenty ten,
there was this thing called machine learning, and we didn't
know what it was at the time, but we said,
you know, it sounds new, so why don't we at
least sort of suss it out and see whether or
not it works. It didn't work very well at the time,
It works really well right now. So it turned out
(11:59):
to be a good decision that we allocated resources because
we got in sort of at that early stage during
the evolution of AI and machine learning. Yeah.
Speaker 1 (12:07):
No, that's interesting. And before actually, before we jump into that,
one of the things I heard from you is it
sounds like there's you the doctors would tend to be
on the safer side, so there would be tend to
be more false positives when they're doing their kind of
subjective eyeball analysis, right.
Speaker 2 (12:22):
Yes, and that's been proven time and time again. Like
we've seen in virtually everything, people tend to air on
the side of conservatism. So what is that I'd rather
call something positive and be wrong than call something negative
and have a catastrophic event happen to the patient.
Speaker 1 (12:39):
Yeah, no, that's it. Yeah, And that's I think that's
why having something that's more objective is very important for
these type of scans that's also done in a timely manner.
Speaker 2 (12:48):
The problem with the false positives that we have for
these kinds of tests, whether it's stress tests or cctas
or whatnot, is that you can air on the side
of conservatism. But by calling it a false positive, what
you do is you end up sending patients to the
next downstream test, which happens to be a very invasive,
very expensive procedure. So that just caused a lot of
(13:11):
unnecessary harm to the patient and a lot of unnecessary
waste to the healthcare system. So we're trying to prevent
all of that.
Speaker 1 (13:19):
Yeah, you rather well measure twice and cut ones type
of c scenario actly. But that kind of gets to
that machine learning because you're you're you, you nailed the
hammer on the nail there where the machine I've heard
machine learning, I feel like since I first started in
the early Yeah you almost ten plus years ago now,
(13:41):
but it never really caught wind until the rise of AI.
And then when you talked about how you started clearly
eight years ago, that was half a decade before all this,
you know, the huge AI bump that we were seeing
over the last two three years. So what were your
initial feedback of that machine learning and kind of what
(14:04):
were the challenges and what were the steps that you
needed to do to kind of overcome those you know,
those challenges.
Speaker 2 (14:11):
The first thing we had to do was learn it,
and so that was pretty tough. Like, we had a
technical computational biology group where I was and you know,
it probably took us a year and a half to
just speak the same language. Right when I would say
something clinical or they would say something technical. We sort
of missed each other for about the first year, year
(14:32):
and a half. Then somehow it clicked where we could
understand they understood the clinical problem to be solved, I
understood the technical approach that they were taking, and so on.
I mean, we've seen iterations of this. I think when
we started the company, that was when the whole sort
of deep learning convolutional neural nets were there that was
good enough, right, Like that was a huge step forward
(14:53):
in terms of image processing to higher accuracy and efficiency.
So we sort of luck doubt at the time because
if we were going with traditional image processing probably wouldn't
the solution wouldn't be half as good as it is today.
And now things have improved and they continue to improve. Also,
what also improves is like the corpus of data that
(15:15):
we can train these algorithms on and then validate them.
You know, we've increased those over time, so we've got
hundreds of thousands of cases that we can access to
really train models in a way that they can recognize
any type of heart disease and really try to offer
this kind of personalized approach to preventing heart attacks. So
we try to support the doctors that way.
Speaker 1 (15:36):
Yeah, and then maybe let's just dive a little deeper
into clearly itself, what the product is, what it does,
how the algorithms work. And then what's interesting is that
you've expanded the offering and so there's now and then
if I understand it correctly, there's clearly, but there's also
clearly a schemia, and there's clearly plaque analysis, and it's
(15:56):
just kind of expanded out. So how is the whole
repertul are of products work, either alone or together to
help you with the patient scans?
Speaker 2 (16:08):
Yeah, it's a good question. Like it goes back to
sort of what we talked about earlier. Again, we wanted
to just make a comprehensive care management platform. What does
that mean? That means that in a single user experience,
I as the user, can get everything that I want
and everything that I need to best take care of
that patient. So who is the definition of I. Well,
(16:29):
there's many eyes, right, there's the imager or the radiologists
who has to read the images. There's the general cardiologists
who've may be taking care of the patient. There's the
preventive cardiologists who more focused on sort of personalizing assessment
and treatment of anthroscrosis. There's the interventional cardiologists who takes
people to the invasive lab and puts in stents. There's
(16:51):
the researchers, there's the primary care physicians and the advanced
practice providers, and most importantly, there's the patient. So the
way that we construc DIDUD this was the first thing
that we made was the care management platform. So if
we're a software as a service company, you just log
onto any onto the Internet from any computer and you
can access this platform that gives you exactly the data
(17:14):
that you need as the stakeholder. So the interventional cardiologists
needs something different than the radiologists. The radiologists need something
different than the primary care doctor, and so on and
so on. So this is the platform that was the
overall structure, and it comes in a way that translates
all of this fancy imaging stuff and pixels into stuff
that doctors can actually understand and patients can actually understand.
(17:37):
So that what we felt was a very important part
of the platform, the translation step, because when we were
at Cornell and New York Presbyterian, we would talk to
all of our colleagues about all this fancy imaging findings
and they're like, I don't know what you're talking about,
because they didn't They shouldn't, right, They weren't imagers. They
were clinicians in the office taking care of patients. So
we realized that all of this important data was getting
(18:00):
lost in the handoff, and so we wanted to rescue
that data by creating this platform. And then after that
we have three additional so we have four fully FDA
cleared products. The first one that one is the care
management platform and the experience, and then the other three
ones are The first one is called Clearly Labs. We
call it Clearly Labs. And what that looks at is
(18:22):
the plaque that is building up very comprehensively in every artery,
in every vessel, in every segment and every lesion. We
can characterize quantitatively all of the differences of the plaque
that's silently building up in the walls of the artery
and putting the patient at risk of a heart attack.
We can also calculate the narrowings in the arteries that
are caused by that plaque. That's called stenosis or blockage
(18:46):
or narrowing. And then we have a second product that
takes all of that information and leverages end to end
artificial intelligence to determine the hemodynamic significance of those lesions.
What do I mean by humanemics significance is those things
cause a reduction in blood flow when say somebody's exercising
or walking up the stairs. That may be the cause
(19:07):
of somebody's chest pain or shortness of breath. So then
what you get are three things. You get this athriskrosis
or plaque that's building up in the wall. You get
the stenosis or the narrowing caused by the plaque, and
then you get whether or not that stenosis is actually
impeding flow on blood flow in a way that may
cause somebody's symptoms. So those are the three major categories.
(19:30):
Within those categories are a ton of subcategories, but the
three major categories of actionable findings for cornerary artery disease
where you know how to act upon it to try
to improve a patient's outcome. And then the last product
that we have really fits into that care management pathway
that we wanted to do around evaluation, education, treatment, and tracking.
(19:52):
It's a quantitative tracking software that we call clearly Compare,
So in the case that somebody had two scans, we
can tell you what's changed in that patient's heart over time.
That's very important because when we give somebody a medication
like a statin, for example, to lower their cholesterol, what
we see is a twenty percent reduction in heart attack events.
(20:15):
That sounds great, except that means that you've lowered cholesterol
in one hundred people and eighty percent of them aren't
benefiting from the drug. We call that population the residual
risk population. But I don't know who they are. I've
given them the drug, the cholesterol is lower, they're seemingly better,
and yet four out of five people aren't benefiting from that.
(20:36):
We think that quantifying disease changes over time will be
could be a very effective way to figure out who
those people are who are at residual risk. And then
it's just a good way to see whether or not
the medicines are working or not, or whether or not
disease is still progressing.
Speaker 1 (20:52):
No, it makes sense because you want to beform an
after and you want to see how the intervention is
actually changed or help the patient's heart. Let me just
ask because it just came to my mind. So you
have all these different products. Are they all packaged into
one commercialize product that you sell to the hospital or
(21:14):
to the radiologists or is there a kind of ALEC
heart menu for these customers to pick and choose which
ones work best for them and their and their hospital budget.
Speaker 2 (21:26):
It's a great question. I mean, it depends on the client.
We you know, I favor sort of all for one
kind of thing, Like I think that comprehensive care is
better than fragmented care, standardized care is better than non
standardized care. So if it were me, I want everything
I want to know about my heart. Some sites adopt
(21:48):
that approach, and we do that through a subscription model.
Others do not, and they would prefer to go all
a cart. We'll meet the client wherever their needs are.
The one thing you know, as you well know, is
you know, for the site to get paid, they typically
have to go through some insurance coverage and reimbursement. And
(22:08):
so we have two codes right now that they're called
CPT codes or current Procedural Technology codes. That's how the
site can build the insurance company so that the patient
doesn't have to pay out of pocket. And so sometimes
depending on how they build and how they submit insurance
claims and so on, sometimes the all the cards better
(22:28):
for them. Sometimes the all for one is better for them.
We just try to work with them to maximize delivery
to as many at risk patients as we can.
Speaker 1 (22:37):
Yeah, and you know, one of the things, you know,
we'll get into reimbursement a little bit more because I
know it's a very evolving landscape in the CMS space.
But I'm curious as well as you know, you talk
about reaching as many patients as possible. There's a plethora
of clinical trials out there teaching, you know, studying different
(22:58):
patient cohorts. And I know one of the most recent
ones is confirmed too, so we can start with that one.
But what are some of the clinical trials clinicals results
when you have the product when it first came out,
that are kind of the important Yeah, study points for
(23:19):
this technology.
Speaker 2 (23:21):
Yeah, it's a great question. There was a we sort
of follow a roadmap. That roadmap is like does this
product work? That's the first question we ask, And so
that's usually a diagnostic performance or diagnostic accuracy question. So
what we've done is we've tested our tools against every
quote unquote gold standard that we can possibly think of.
(23:44):
That turns out to be the invasive methods of imaging,
which have you know, they're closer to the hard arteries
because they're inside the arteries and so they have very
high resolution. So we've tested against five different invasive modalities.
I won't go in to all of the names of them,
but we do that with blinded core laboratories against these
(24:05):
gold standards and have demonstrated high performance. The second question
that we ask is is what I'm detecting even relevant
doesn't matter and there are tests out there that if
they're normal or they're abnormal, it doesn't really affect your life.
So then the question is why are you measuring it?
(24:25):
Because if it's not prognostically useful, then it's not a
particularly useful test. And so we've done a series of
large scale studies with two, three, four, five, seven, eight,
and ten year outcomes to demonstrate that when we see something,
we can identify people at risk in a very graded
fashion so that we understand who needs to be treated
(24:48):
more aggressively. The third thing that we've done is a
series of large scale clinical studies that have eight, seven
and eight year follow up to make sure that when
people are acting upon our study findings that the patient's
outcome can actually be improved, that we can reduce future
major adverse cardiovascar events, things like heart attacks, death stroke,
(25:09):
and things like that. So we've just presented two large
scale studies at the American College of Cardiology meeting a
few months ago. The fourth thing that we do is
to try, you know, we should be stewards good stewards
for the healthcare system is extremely big spend for the
US government and for patient families that are paying commercial insurance,
(25:32):
and so we want to make sure that our tools
don't just add bloat to the healthcare system, but actually
have the potential to actually significantly bend the cost curve downwards.
So we've spent a significant amount of time publishing some
studies that demonstrate cost savings associated with our tools as well,
and so those that's sort of the tiered approach that
(25:55):
we've taken to try to prove our stuff. Coming back
to what you asked about, we believe like all of
the stuff that went into clearly was based on science, right,
all these large scale clinical trials. So that's our DNA, right.
We ran clinical trials for twenty years, and so we
are committed to continuing that science evolution and sort of
(26:21):
searching for always the right answers, and we will continue
to learn for the rest of our lives, because that's
just the iterative process of complex human biology. But the
easiest way you can learn that is from very high quality,
large scale studies. And so the one that you suggested
(26:41):
was Confirmed two. The reason it's called Confirm two is
because when I was at Cornell, we did Confirm one.
That was a study of twenty three thousand patients who
we had followed for two years. Confirmed two goes it's
like Confirm one on steroids. So what we have done
is now we've amassed fifty some sites across the globe
(27:05):
from multiple different ethnicities and countries, and what we have
done is about five year follow up of these patients
who have undergone everything that clearly has to offer, so
advanced plaque analysis, advanced narrowing analysis, advance the schemia analysis,
and we've adjudicated all of the outcomes, and so it's
(27:26):
just teaching us a lot about not only clearly tests
and how that impacts somebody's outcome, but also just Vasker
biology in general. We're learning brand new biological concepts. As
one of them we presented at the late breaking clinical
trials for the American College of Cardology was gender differences.
We fully don't fully understand the difference between women and men,
(27:49):
but when you look at the data, Wow, women and
men behave extremely differently when it comes to heart disease,
and we need to recognize that in order to better
take care of women who we've sort of lumped into
the same category as men for too many years. And
so it's just that one is we'll advance our understanding
of science, and then it's really going to validate the performance,
(28:12):
the prognostic performance of our products in many, many studies
to come.
Speaker 1 (28:18):
Yeah, and that's the thing about confirmed too that caught
my attention is it was a fifty to fifty split
between men and women, when usually these clinical trials are
have the women are well underrepresented based on you know,
relative to kind of the actual treatment options that are available.
What caught my attention to as well. And actually we'll
probably cover that with the reimbursement side, but The other
(28:43):
one too that's coming out that's currently enrolling, is the
Paramount trial. And what caught my attention is that's a
randomized trial. So what's the importance of doing that to
help with you know, making you know, CT scans and
the AI funk features with these CT scans become more
(29:04):
prevalent at these hospitals.
Speaker 2 (29:05):
Yeah, it's a great question. So we have two large
scale randomized control trials ongoing. One of them is Paramount,
the other one is called Transform, and I'll tell you
the rationale behind both of them. The you know, for
the listeners, like, a randomized trial is where you take
a population of people and you subject in general, half
(29:26):
to one thing and half to another. And that's to
just alter one variable in the experiment and then to
see what happens and what changes in the outcome. That's
the sinaquonon for cardiovascar medicine. Every you know, we cardiology
is the most evidence based field and it has relied
on the mechanism of the randomized trial. And it's because
(29:49):
it's a very controlled That's why it's called a randomized
controlled trial. You're just changing one variable and you see
what the outcome is. So what's the difference between that
and say, like a confirmed to confirm two is an
observational cohort study that means you just take ten thousand
consecutive people and you watch them, and that's called a registry.
So the randomized trial is great for proving one thing,
(30:13):
but there's always inclusion criteria and exclusion criteria, so it's
not everybody. So to me, what randomized controlled trials does
is proves your point and really gets you guideline quality
data to change the American Heart Association American Culture Cardiology guidelines.
And what registries do is it teaches you how to
use it in clinical practice, because those are all comers,
(30:36):
and so we try to do both of those. So
for the two randomized trials that we're doing, one of
them is called Paramount. It is a randomized trial of
patients who have symptoms suggestive of heart disease, chest pain,
shortness of breath, and so on. And what we wanted
to do was really look at the influence of what
clear these technologies could do to improve patient health outcomes,
(30:58):
to reduce seri invasive procedures, to improve medical or cornority
disease resk fact or control, to really show that like
this is good for patients, it's good for the healthcare
system at large. And so yeah, that trial, we're aiming
to finish enrollment by the end of this calendar year,
and that will hopefully prove the role of end to
(31:21):
end AI within a single, unified, comprehensive care management platform
to improve patient health outcomes. The second trial is called Transform.
It's a much bigger swing at the bat. So if
you look and you asked me, what is the biggest
unmet need in coronary heart disease the cause of heart attacks,
(31:41):
I will tell you it is the asymptomatic people walking
on the street. And you'll scratch your head and say,
I don't understand that. Well, it turns out everybody thinks
of heart attacks as having chest pain and so on.
That's very common, and everybody thinks that the precursor to
having a heart attack is having chest pain for a while,
and then you don't you know. But what happens when
you look at the data is that somewhere between fifty
(32:04):
and sixty percent of people who will die from a
heart attack or suffer a heart attack never have a
symptom before their event. And I think you and I
know everybody knows someone like that, right who went for
a run and died and never came back, or went
to sleep and never woke up. And the reason everybody
knows something like that is that it's the number one
(32:24):
cause of death in this world. And I can be
the best preventive cardiologist in the world in you know,
at New York Presbyterian Hospital or wherever, but those patients
don't come to me because they die in the home.
So the biggest unmet need is to find those asymptomatic
people who comprise the majority of people who will have
heart attacks and go find them. And I think the
(32:45):
only way you can do it is to go up
into the home. And so we have this trial called TRANSFORM,
a randomized control trial for screening completely asymptomatic individuals. They're
not patients, they're not people in the hospital. They're the
people that you see walking in the grocery restore and
playing Little League on the weekends who then die suddenly
of heart attacks. And so what we're it's a seventy
(33:07):
five hundred patient study. We're about thirty seven thirty eight
hundred patients enrolled, and we are testing the hypothesis that
imaging actual disease of a patient is better than checking
risk factors of disease like cholesterol. So you know, we're
we're still we're enrolling as quickly as we can because
(33:29):
I think this is the number one public health epidemic.
Like in twenty twenty, there are twice as many cardibaskard
desks that than desks from COVID nineteen. That's how big
this epidemic is. And so we feel very motivated that
there's an urgency to us to find the answer to
better take care of patients.
Speaker 1 (33:44):
There's a lot to unwrapped there, because you're totally right,
and you thought that, you know, I would be surprised
when I hear about you know, fifty sixty percent, you know,
don't have any symptoms before. But you hear countless stories
either at a personal level or you know, just in
the news and things like that, and then all you
just hear is like, oh, they just dropped dead. They
were perfectly fine, you know, as you said, they were running,
(34:05):
and then just it's unfortunately, you're right, it's so it's
so tragic about how instant that could be.
Speaker 2 (34:12):
Just a short story, Like two weeks ago, I was
talking with a doctor and she had ordered a number
of clearly analyzes on her patients, and she was just
showing me, like what she one patient, Like she had
deidentified the patient, but she said that he was pretty young,
he was in his mid fifties, and just showed me
(34:33):
the clearly analysis findings and they were pretty marked, and
I thought, wow, this person is quite sick. Like what
did you do with this patient? And she said, well,
I recommended that he go on certain medical therapy and
I said, yeah, I think that. I mean, just as
a former practicing cardiologist, that sounds pretty reasonable. And I
said what happened? She said, he refused it. He said
(34:54):
he didn't want to take any medications. So they agreed
that they would talk about it the following week and
just to see what the plan should be. He died
on Monday, like four days later. So it's like those
kinds of things that just make you so humble that like,
we really need to tackle this problem. It's and that
person was so young in their mid fifties, Like there's
(35:15):
just too much of that. So that was our north
star when we started the company, that the tool we
wanted to make was a comprehensive care management platform, But
our north star was really to try to accrue the
evidence and prove or disprove the utility of just measuring
disease over risk factors of disease to prevent heart attacks.
Speaker 1 (35:35):
Yeah, and let's just let's dive into that because I
know in the past you've mentioned that you would want
to imagine a world where there's no sun in death
from heart attacks, and here we are just talking about
those instant, tragic events. But there's a lot of steps
that need to get there, and well, maybe you can
just jump into each of those steps. But one of
(35:56):
the first ones that come to mind is the cost
of a CT scan versus some of the other traditional
methods of scanning for patients. And there's also the reimbursement landscape.
So walk us through. You know, the cost difference is
the reimbursement because, like you said, sometimes the hospital gets
a certain reimbursement and they only can only do so
(36:18):
many scans, they only do one or two of your
comprehensive skins. And then talk about the you know, costs,
economics and the data that you're seeing that could say, hey,
it might be a little bit more now, but in
the long run, this is the right decision to make.
Speaker 2 (36:37):
Yeah, it's both of them are really great questions. So
let me be as clear as I can. We are
not advocating right now for universal screening at all. And
the coverage and reimbursement that we have for our tools
is solely restricted for symptomatic patients with suspected heart disease,
shortness of bread, chest pain, and so on, and that
(36:59):
is currently where our company is focusing its commercial efforts
on that symptomatic population. So I don't want it. I
want it to be very clear that we are not
here advocating for universal screening until or unless this large scale,
randomized trial is positive, and at that point in time,
then we'll start to lobby all of the insurance providers
(37:19):
to say, look, we can save lives here. So for
the first one, that's the symptomatic population, I think that's
a traditional fee for service model. What's that mean? Like
I do a service, you give me a fee, the
insurance company gives me if the hospital fee and so on.
There's also value based models that we I think fit
very well in where we can be a better personalized
(37:42):
referral management tool to reduce costs on the screening side.
Which is I think where you were directing some of this.
You know, there's a number of ways that you can
approach this. They're all long, they're all very expensive, and
they all require level grade A science. And if you
don't have that, then I don't think you have a
prayer to try to move this kind of movement forward,
(38:05):
particularly as big a frameshipt as we're trying to espouse.
When you look at the data though, and you look
at the estimates, when we said, okay, what are we
going to see in our clinical trial, you have to
come up with a hypothesis we're going to reduce heart
attacks by ten percent, twenty percent, thirty percent. And so
when you look at that, we talk about a number
(38:25):
needed to screen in order to identify and prevent a
bad thing from happening. For mammograms, the data that's been
published is that you need about six hundred women to
undergo fifteen consecutive years of mammograms in order to prevent
one breast cancer death. For colonoscopies, it's around thirteen hundred
(38:46):
and fifty patients if our number needed, If our hypothesis
is correct for this trial, the number needed to screen
would be thirty so if you believe that screening mammograms
and screening colonoscopy are cost effective and that we should
be doing them, then this should be wildly cost effective
(39:07):
if the trial is actually positive, right, and then in
terms of the price, fortunately or unfortunately, the cordinerary CT
indrogram is not particularly expensive, Like it's one of the
lower cost advanced imaging tests out there. So all in,
if you were to bake in the CT scan costs
(39:28):
and the clearly analyzes certainly be less than a colonoscopy
which we're using for universal screening, and then it would
identify a lot more people at risk. So you know,
we have to finish the trial. We have built in
very complex advanced cost effectiveness analysis within the trial, so
we'll we'll provide all of that data as well. But
(39:50):
there's just too many people dying of heart attacks, Like
we've got to do something. And when you said, well,
where's the problem, I see three different legs of this
stool identification, treatment and implementation. If you look at the
treatment like just currently what we have today in our toolbox,
which is about a dozen different medications and lifestyle modifications.
(40:12):
If you do the additive relative risk reductions over ninety percent,
we can eradicate heart attacks. What we don't do is
identify well. I think that's where it clearly really comes
in to hone in on a precision analysis of somebody's heart,
and then we don't implement well because we wait for
people to come to the hospital in the throes of
some catastrophic event rather than to try to prevent it
(40:34):
early when it can be much more easily prevented. So
this whole like we're trying to address those two legs
of the stool, the identification and the implementation.
Speaker 1 (40:44):
Yeah and yeah, it seems like you're in your adding
the to RCT, using especially transform is definitely the right
step because you're right, sometimes when you wait for the symptoms,
it's too late at that point. So it sounds like though,
I mean, right now you're advocating for only symptomatic because
you're waiting until the RCT results come out and it
(41:06):
hopefully supports your hypothesis to start training the sooner if
everything goes as you and vision, How do you see
this technology being utilized over the next five to ten years,
And especially what caught my attention to is you talked
about iterative not only for the data for the AI
tools to keep learning, but also the human aspect of it.
(41:29):
As you were saying, you're always going to continue to learn.
How do combining both of those segments together create what
AI truly is designed to help society out with?
Speaker 2 (41:43):
I mean, like going forward. I can answer that in
a number of different ways. Like, we have a current
set product roadmap that's over ten years long, so we're
nowhere near done. It will likely change over time because
we'll be influenced by new and inovations that come about.
And what's embedded what's not embedded in that product roadmap
(42:07):
is sort of the power of what AI can offer,
which is inferences from data and data at scale. Right,
So you know, I believe for heart disease that seeing
heart disease, actual heart disease is the most effective way
to identify somebody who is sick and to guide your treatment.
I believe that in my heart of hearts. That doesn't
(42:30):
mean that other data is not important. Right. So there's
plenty of particularly in the cancer space. There's these liquid
biopsy companies, right, and some look at genomics and polygenic
risk scores, some look at epigenetics, some look at proteomics,
some try to combine them altogether and so on and
so on. Or there are companies out there who are
focusing on biometrics that you get from wearables, your rr
(42:52):
A ring or your Apple watch or your garment. I
think they're all useful, I think to a lesser degree
than whether you actually have heart disease. But I think
the future of AI is that we will integrate all
of these seemingly disparate data and we will output it
in a way that would be clinically actionable. To that
(43:12):
you just need a ton of data, right, Like that
means you need blood samples and you know what your
watch said and did you sleep well, and what's your whatever.
So like that's going to be take some time to
approve that kind of really high quality data, and then
it's also going to require a very respectful approach to
(43:33):
regulatory approval and to coverage and reimbursement, right, So those
are things, and then it's got to be proven a
science and proven through large scale clinical trials. So like,
I don't ever worry that AI is going to outpace us,
because I think AI is just a tool. It's a
(43:55):
tool that's embedded into a medical device, and because it's
a medical device. It requires the same careful, meticulous approach
towards regulatory approval, towards evidence generation, towards large scale clinical trials.
And so I don't think the field of medicine ever
will or should outpace the AI technology improvements because the
(44:17):
stakes are too high. Right, the AI gets it wrong,
somebody dies, that's just not acceptable. So you have to
just keep testing it in an evidence based fashion to
prove beyond certainty that the device works and it actually
improves patient healthcare.
Speaker 1 (44:33):
Yeah, and you know what, at the end of the
day that all works out. That's how you optimize AI
in this world. But Jim, thank you so much for
joining us today. It was a very enlightening episode.
Speaker 2 (44:43):
No, it's a pleasure to be here, and thanks so
much for having me on the show.
Speaker 1 (44:47):
Yes, and thank you to our listeners for tuning in today.
We hope you join us for future episodes, and if
you'd like to stay up to date, you can click
the subscribe button on Spotify or your favorite streaming platform.
Take care Chasser uses US
Speaker 2 (45:24):
Changes bases
Host
Jonathan Palmer
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
The Joe Rogan Experience
The official podcast of comedian Joe Rogan.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Special Summer Offer: Exclusively on Apple Podcasts, try our Dateline Premium subscription completely free for one month! With Dateline Premium, you get every episode ad-free plus exclusive bonus content.