Jun 1, 2021
In the 3rd part of our 3-Part Series on Ways to Attack Pulmonary Vascular Disease, Stanford Pulmonary Hypertension specialists Drs. Vinicio de Jesus Perez, Edda Spiekerkoetter & Andrew Sweatt discuss the adult clinical approach and ways the Wall Center is fighting pulmonary hypertension across multiple fronts.
Vinicio de Jesus Perez, MD
Edda Spiekerkoetter, MD
Andrew John Sweatt, MD
Welcome to the PH at Stanford Podcast. This new podcast series comes to you from the Vera Moulton Wall Center for Pulmonary Vascular Disease at Stanford, with the goal to eradicate pulmonary vascular disease by discovering fundamental causes, developing innovative therapies, disseminating crucial knowledge, and delivering transformative care.
Today, is the 3rd in a three-part COVID-related series on Ways to Attack Pulmonary Vascular Disease. Stanford PH Drs. Vinicio de Jesus Perez, Edda Spiekerkoetter, and Andrew John Sweatt, discuss the adult clinical approach and ways the Wall Center is fighting pulmonary hypertension across multiple fronts.
Edda Spiekerkoetter, MD:
My name is Dr. Edda Spiekerkoetter. I'm an adult pulmonologist at Stanford, and I'm treating patients with pulmonary hypertension and also have a basic research lab where I study the pathogenesis of pulmonary hypertension, right heart failure and vascular malformations in the lung.
Andrew John Sweatt, MD:
Hi, I'm Dr. Andrew Sweatt. I'm also in the division of pulmonary and critical care at Stanford, treating adult patients with pulmonary hypertension and other general pulmonary conditions. And my clinical research focus is in pulmonary hypertension.
Vinicio de Jesus Perez, MD:
My name is Dr. Vinicio de Jesus Perez. Like my colleagues, I'm an adult pulmonary critical care specialists and physician scientist with expertise in pulmonary arterial hypertension and other pulmonary vascular disorders.
Today, we're going to talk to you about what the Stanford Pulmonary Hypertension Program is doing to fight pulmonary hypertension across three fronts. Number one, the repurposing of drugs, discovery of biomarker through omics technologies and the use of wearables for monitoring pulmonary hypertension patients.
I will start first by telling you about our work with wearables. This is work that has been done in partnership with our friends at phaware [global association], and it is a tool that will combine the use of the technology in the Apple watch together with an application developed with our colleagues at phaware, that will allow our patients to capture six minute walk data from the comfort of their living place, whether it's their home or their local park.
Now, why is this important? Well, many of us have been affected by the COVID epidemic. As you all know, when you visit your physician, one of the key tests that the physician will offer is the six minute walk test. Why? It tells us how well your heart and lungs are working together. It also tells us whether the medications and interventions that we are offering you are having the expected impact.
The problem being that with COVID is it has become incredibly hard to host patients in our clinic. The six minute walk test, is traditionally a test that has to be done in clinic under the supervision of a respiratory therapist and with a proper assessment of the pulmonary hypertension specialist. For many of us it's become very difficult to gauge how our patients are doing through video consultation in regards to the level of activity and how their medications are actually influencing that.
To overcome that barrier, phaware and our group have partnered to test this new app, which is the phaware Walk.Talk.Track™. This is actually an app that works with the Apple watch, and it links wirelessly with your Apple phone. What the app does is it actually instructs you on how to capture a six minute walk test from the comfort of your home or any place locally where you can actually walk. It's simple. The app allows you to initiate the test, capture your symptoms, and then it tells you to walk for six minutes. It tells you when to rest. Once you're done, that information is uploaded to a cloud. That information comes to us using sophisticated algorithms that we have been developing. We can capture these numbers, your symptoms at rest, how much you walk, your heart rate, the changes in your heart rate. And, this information is captured on a daily basis, or as frequently as you can do.
Now, why is that important? Because I'm not depending only on me seeing you every three months in order to know whether your functional capacity is improving or not, or maybe worsening. In an ideal situation where I have access to this technology, I can potentially identify a level of worsening that will prompt me to introduce a therapeutic intervention. For you, it means waiting less time in order to get the treatment that you require so that we can keep your quality of life and your level of function high, which is something that you will appreciate, your loved ones will appreciate, and for us, it will make our jobs much easier. This is currently being tested. We recently had two successful presentations at the American Thoracic Society that were featured as part of the program that I invite you all to check.
We're currently conducting these studies. We hope to be able to put this out very soon. Once we publish and validate this, this will be available free to all users. And as I said, this is thanks to the support provided by phaware. We're very proud to be part of this incredible project. I'm hoping that this is of great interest to you. Now I'm going to pass the baton, if you will, to my colleague Dr. Spiekerkoetter, who will tell you about drug repurposing in pulmonary hypertension.
Edda Spiekerkoetter, MD:
Thank you so much, Vinicio, for this introduction. I would like to switch gears a little bit and talk to you about what Stanford does to find better treatments for pulmonary hypertension. So to start with, I'll give you a little bit of an overview about pulmonary hypertension. Pulmonary hypertension is characterized by narrowing of the lung vessels, in particular the pulmonary arteries. This is due to cell growth in the vessels, which we also call neointima formation. Many of the current approved drugs that we have mainly dilate the pulmonary arteries, but they don't affect the cell growth in the vessels. The disease often progresses, despite treatments and patients at some point worsen, develop heart failure and some even need to be transplanted, and receive a new lung. What we're trying to do is finding new drugs that could really reverse the cell growth in the lung vessels to reduce the resistance and make it easier for the heart to pump blood through the lungs. A way that we have been trying to achieve this to repurpose or reposition drugs used for other diseases.
Repurposing of drugs means that you use drugs that are not approved yet by the FDA, but that have been tested by pharmaceutical companies yet that have failed for a certain disease. It is possible to use these drugs for a different purpose and see for example whether they work in pulmonary hypertension. One drug, for example, that we're actively working on is the repurposed Lymphoma drug Enzastaurin. I will tell you a little bit more about this later. Repositioning of drugs means that a drug already FDA approved for a different disease or indication is used for a new indication. By doing so, we're using off-target effects or side effects of the drug to see whether they might be beneficial for pulmonary hypertension and can improve the disease. One example of a repositioned drug that we are studying is the immunnosuppressive drug FK506, or Tacrolimus. Often the terms repositioning and repurposing are used interchangeably--and generally mean using "old drugs" for new indications.
In pulmonary hypertension, many drugs have been repurposed over the years and maybe probably the best drug that you know is Sildenafil. Sildenafil, also called Viagra, was originally developed to treat angina chest pain because it was known that it relaxes smooth muscle cells. Then it was found as a side effect that men develop erections and so that's when some smart people had the idea to say, well, let's use this as a good drug for erectile dysfunction. Then it was repurposed again for the use of pulmonary hypertension, knowing that, well, if this dilates vessels, relaxes vessels, it might even help pulmonary hypertension. That's how we got the drug Sildenafil for PH, which many of you guys are using for pulmonary hypertension at a lower level than it was originally developed for.
The advantages of drug repurposing is actually that it's much
more cost effective than if you develop drugs de novo. It's also
much faster because new drug development often takes 12 to 15
years, whereas when you repurpose drugs, you can skip many of the
different steps in drug development: The toxicology is known,
the dosing as well as the side effects are known. Therefore, you
can really start testing promising drugs in patients and perform
Phase II clinical trials in patients early on, without having to
spend time with lengthy preclinical testing.
Because the side effects are known, it's often possible that repurposed drugs can be used at a lower dose. This has many advantages for the treatment for pulmonary hypertension. The downside is that the repurposed drugs are often generic. Many pharmaceutical companies are hesitant or less interested in funding larger trials. That's where for example, centers like the Vera Moulton Wall Center for Pulmonary Vascular Disease or the National Institutes of Health or other funding opportunities play a big role to help fund those clinical trials and make these drugs available and test them in patients.
Stanford has a long history of performing what we call investigator-initiated clinical trials. We test promising drugs in very focused small trials with our patients. And I'm sure as a patient at Stanford, you have been asked to participate in several of these trials. We recently completed a trial with the drug FK506 or also called Tacrolimus, which is an immunosuppressive drug that has been used over 30 years in patients after organ transplantation. We tested, whether by activating the BMPR2 pathway that is downregulated in pulmonary hypertension, FK506 reverses the cell growth in the lung vessels, reduced neointima formation and improved pulmonary hypertension. As a second trial we are currently applying for funding at the NIH for a multicenter trial to test whether the cancer drug Enzastaurin, improves pulmonary hypertension. A pharmaceutical company offered to supply the drug for free.
In the past, it was often that people just looked at side effects of these drugs and then felt okay, if certain drugs have these and these side effects, maybe we can use the side effects for our disease as a beneficial effect. But more and more, we are also doing experimental screening. So high throughput screening of drugs to determine whether certain drugs activate a specific pathway, and this was actually how we found, FK506. More and more, we are also using computational approaches now to predict which drugs might affect certain pathways. That's where bioinformatics, artificial intelligence and drug prediction comes in. So I would like to hand the baton over to my colleague, Dr. Andy Sweatt, who will talk a little bit more about this topic.
Andrew John Sweatt, MD:
Thank you, Edda, for the introduction and the segue into what I'll be talking about, which is how we can use and how we've been using artificial intelligence and machine learning to help yield new insights and discoveries in pulmonary hypertension specifically. I think these terms, artificial intelligence and machine learning, are something that people are gaining increasing exposure to over time in the popular media and culture. It's commonly talked about on certain commercials.
We know artificial intelligence has led to self-driving cars, and it's the reason that all the social media platforms are able to tailor advertisements specifically to our individual interests--in interest of trying to get more clicks as well as facial recognition software that's embedded in smartphones. These are only a few common examples of artificial intelligence, which is essentially a field of computer science that aims to mimic human thought processes, learning capacity and knowledge storage to provide useful tools.
Machine learning is actually just a subset of the field within the larger umbrella of artificial intelligence. Machine learning is what I'll be talking mainly about today, which is the use of computational algorithms and statistical models to find hidden patterns in complex datasets that would otherwise be undiscovered, or to make predictions or identify patterns among variables within this complex dataset.
Before I talk more, I think it's important to understand that machine learning approaches can be broadly split into two main categories. Those that are supervised machine learning. These are methods in which you're trying to predict something such as predict a certain outcome or classify or predict a certain feature, such as is this a chicken or a duck, if given a picture and then the machine algorithm will spit out the prediction this is a chicken, this is a duck. Whereas unsupervised machine learning is really designed to divide up a dataset into new subsets within that data set, identifying novel subgroups of something.
For example, in pulmonary hypertension, the use of all machine learning approaches is kind of recent on the stage. Our group has been one leader in this field, but there's other groups also increasingly applying these approaches to help us gain insight. Potential applications in pulmonary hypertension are very, very broad. When it comes to supervised approaches, this is when you have a dataset, if you want it to predict pulmonary hypertension outcomes like survival or something, and develop a tool that can predict how a patient is going to do over time. That's one way to do it.
For example, you could use data from the wearables that Vinicio talked about before, the six minute walk distances over time. We may be able to identify an early signal with the computational algorithms and predict who's at risk for hospitalization or bad outcomes over time. These tools have been used to try to spur early detection of the disease. Pulmonary hypertension, the diagnosis really relies on an invasive procedure as we know right now. So the idea is to use noninvasive data ranging from echo metrics, imaging metrics, or even just claims-based data in diagnostic codes has been used. These have been applied to try to identify who's a pulmonary hypertension patient, and who's not to help diagnosis. More specifically, my interests have been to use machine learning as a way to facilitate what we call deep phenotyping in disease.
My initial interest in this field in general kind of spurred from my desire to combine my undergraduate biomedical engineering and computer science training with my subsequent clinical interests that I gained through medical school, residency and fellowship, which was pulmonary hypertension. I think what I was struck by as I was exposed to my first pulmonary hypertension patients, is that such a wide diverse array of patients developed pulmonary hypertension. It can be due to genetic causes. It can be due to auto-immune diseases. It can be to certain drugs and toxins exposures. Yet they all end up with the same disorder, pulmonary hypertension. Unfortunately, regardless of the cause, we are at the stage in pulmonary hypertension where we treat all the patients the same, kind of a one size fits all approach. This is the way we classify the patients based on the underlying cause.
My initial thought is there must be a more elegant way or there's probably another way to classify patients that would more relate to the underlying biology of the disease and potentially help us lead ourselves and pave a path towards targeting certain subgroups of patients with certain therapies who are more likely to respond to these specific therapies. At Stanford, we have a rich history, long before I was here, in making discoveries that relate to inflammation in pulmonary hypertension, as it relates to the development of the disease process.
My foundational work started with measuring a large number of inflammatory proteins and markers in the blood of pulmonary hypertensions, and then applying an unsupervised machine learning approach to really try to find patterns in this complex data, where you are measuring a large number of proteins at the same time. The algorithm detected that there are these four subtypes of inflammation that exist. Lo and behold, when I looked at the clinical outcomes and characteristics over time, these patients had differences in disease severity and outcomes.
Yet these subgroups that I defined did not correlate with the clinical subtypes, such as idiopathic PAH, connective tissue disease PAH. Essentially, this tool is a way to reclassify patients according to this immune phenotyping approach. Now my work is going to be to take this further and look at what happens to immune phenotypes more over time during the disease stage. I'm going to analyze data from clinical trial secondarily, to understand how these immune phenotypes respond to certain therapies to see if they have differential responses.
On that side, I've been kind of applying unsupervised machine learning, and then more recently I've been applying supervised approaches to identify patients who have adverse right heart effects of pulmonary hypertension over time. And as well as a study led by a clinical trial, NIH funded, led by my colleagues, Dr. Mark Nicolls and Dr. Roham Zamanian of Rituximab an immune modulating agent in PAH, kind of a repurposed therapy that has been used for autoimmune diseases in general, for years, similar to what Edda talked about. This is used for specifically patients with scleroderma associated PAH. We measured a bunch of proteins in kind of exploratory fashion in the blood, inflammatory markers, and I defined a signature when measured at baseline before drug exposure predicts a more favorable response to the therapy. However, this was in one small cohort and this signature is something that would need to be further validated and proven in subsequent groups of patients over time.
We've heard already that PAH is a rare disease. It's time consuming to run clinical trials, because it's harder to find the patients to enroll than in more common diseases like coronary artery disease or run of the mill systemic hypertension. A lot of trials of novel therapies have failed in recent years, and the thought is that we need to be targeting certain subgroups of patients who are more likely to respond to these therapies to begin with. Machine learning may be able to help us achieve that fact.
Before I finish, I'll close with the concept that sometimes there's this concept that machine learning and artificial intelligence technologies are something that are going to, you know, "the robots are going to take over the world" mentality. I think that when it comes to self-driving cars, these are machine learning and artificial intelligence algorithms that have been taken to the max level of sophistication, whereas in the pulmonary hypertension and other medical fields we're using the kind of under-the-hood algorithms, the simple algorithms that form the basis of that to help guide our research. Essentially the important thing is that human oversight is very necessary in this process. People don't understand what leads to the predictions that the algorithms make or this. It's going to be important in our field as we apply these approaches, to be very transparent in our approaches and try to present the data in a way that's interpretable for both clinicians and the PAH community at large.
It's also important to be aware of this concept that that has gotten a lot of attention recently called algorithmic bias, where essentially this is when machine learning models make inaccurate predictions and unfairly disadvantage certain subgroups of individuals due to inherent biases in the way the data you use to train these models is collected, labeled and utilized. For example, if you have disproportionate ethnicity breakdown in your data set, you're more likely to unfairly disadvantage certain subgroups. So there's a whole field of science now focused on trying to avoid algorithmic bias. I think the take home point is, is that there has to be a large collaborative effort in our field in order to implement these technologies and approaches in our field, while still being transparent and having a lot of human oversight in the process.
There's a lot of exciting research happening at Stanford and
with the Vera Moulton Wall Center for Pulmonary Vascular Disease.
We've been able to keep our momentum up despite challenges
presented by the COVID pandemic. We're excited to accelerate
research now that the viral prevalence is slowing down. As you can
see this kind of research that we are all doing would not be
possible without all the patients generously volunteering their
time and effort. We hope that we can continue these efforts moving
Thanks to Drs. de Jesus Perez, Spiekerkoetter and Sweatt. And thank you for joining us here today on the PH at Stanford Podcast. Join us next time.
Until then, you can learn more about the Vera Moulton Wall Center for Pulmonary Vascular Disease at Stanford’s vision to transform the way pulmonary vascular disease is understood and treated, both locally and globally at www.stanfordph.org and be sure to subscribe to the PH at Stanford Podcast on iTunes or wherever you get your podcasts.