About This Episode

In this episode of Dialogues, we continue our look at Translating Science Into Medicine by talking with Dr. Doris Fuertinger, Director of Biomathematical Modeling and Simulation at Fresenius Medical Care, and a pioneer in the concept of virtual clinical trials.  Dr. Fuertinger and her team are spearheading new techniques using complex mathematics and real-life data to create patient Avatars that can help predict an individual’s physiological response to a certain treatment.  These realistic simulations of clinical trials can help overcome the limitations of traditional approaches and lead to new improvements in patient outcomes.

Featured Guest: Doris Fuertinger, MSc, PhD

Doris Fuertinger received her master’s degree in mathematics in 2008 and a doctorate in applied mathematics in 2012. She pioneered the concept of a virtual dialysis clinic and virtual clinical trials, and together with her research team developed methods to apply them to the area of bone mineral metabolism, anemia, and fluid management of dialysis patients. She has authored and co-authored multiple papers and book chapters, and is inventor on a number of international and US patents held by FMC.


Listen to This Episode



Episode Transcript:

Clinical trials play a vital role in the medical and scientific advancement of patient care, but also come with distinct challenges and limitations. In this next episode of Dialogues, we continue our look at Translating Science Into Medicine by talking with Dr. Doris Fuertinger, Director of Biomathematical Modeling and Simulation at Fresenius Medical Care, and a pioneer in the concept of virtual clinical trials. Dr. Fuertinger and her team are spearheading new techniques using complex mathematics and real-life data to create patient Avatars that can help predict an individual’s physiological response to a certain treatment. These realistic simulations of clinical trials can help overcome the limitations of traditional approaches and lead to new improvements in patient outcomes.

Dr. Frank Maddux: I’m here today with Dr. Doris Fuertinger and Doris, thanks so much for joining us today with our dialogue session.

Doris Fuertinger: My pleasure. So, thanks for having me.

Dr. Frank Maddux: I thought we might start by letting you just tell us a little bit about your background and how you got into, not only mathematics, but this kind of applied mathematics that you do.

Doris Fuertinger: My background, actually, is in applied mathematics and my training has been quite technical in the beginning. So, the university I studied at and actually the city I used to live in houses a lot of technical and industrial companies, so, among them, suppliers for the automotive and airplane industry. So, a lot of my colleagues and peers have actually pursued a career in this area, but I very early on was interested in life sciences and I was absolutely fascinated by the idea that you could use mathematics to actually better understand the workings and the functions of our bodies. So, this is where I kind of got hooked with mathematics and life sciences. So, you might think “How can mathematics actually contribute to the advance of medicine?” and well, medicine and mathematics actually are maybe more closely interwoven than many of you might think. So, to give you an example, in the 1700s, the people believed that human blood is entirely built in the liver. So, there’s no circulation. Blood is continuously created in the liver. Now, William Harvey, a physician at that time, used mathematical considerations and some experiments to actually show that the blood is pumped through the heart and that it’s circulating through arteries and veins in our body. So, this example really stuck with me when I first read that.

Dr. Frank Maddux: It’s a fascinating career path and I think we’ve been able to watch you develop this opportunity for us to change the way we look at clinical trials and I wonder if you could just talk a little bit about the challenges of clinical trials and the both potential that we believe and that we’ve actually realized around virtual clinical trials.

Doris Fuertinger: Without doubt, clinical trials and specifically randomized clinical trials have advanced medicine in a great way. However, clinical trials face some limitations that actually limit the generalizability. So, one of the problems we have with clinical trials is that sometimes, the population that we study is not representative of the wider population of interest and this is actually especially tricky in kidney disease patients because in clinical studies, you usually like to exclude elderly patients, frail patients, patients with a lot of comorbidities and one or more of these things applied to many of our patients. Another problem is that we have to enroll a sufficiently large study population. Now, a study on hemodialysis patients sponsored by the NIH actually faced such a challenge. So, they screened more than 6,000 patients and only 240 of them actually ended up being enrolled into the clinical study. Another problem that we have is that you have to build a sophisticated trial structure, infrastructure and this on its own is very challenging and is associated with high costs. We have this really long duration from study inception to study conclusion and one very important thing we should not forget is that in a clinical study, you are limited to test one or a maximum of two or three treatment options at a time. Now, we think that virtual clinical trials are a way out of this predicament, to a certain extent.

Dr. Frank Maddux: So, we’ve got clinical trials that are expensive, slow, and sometimes difficult to execute on and we’ve got mathematicians who have recognized they can begin to model human physiology. So, let’s try to break down the pieces of it. Tell us just a little bit about how you go about mathematically modeling a human physiologic pathway.

Doris Fuertinger: I might actually have to go back even a little bit further to my background and my education. So, in the beginning, I mentioned that my background is quite technical and this, to a certain degree, has actually shaped my way of thinking. So, when we-- at least bear with me here a little bit-- when we look at the automotive industry, the automotive industry came from testing the safety of the cars in animals and then progressed to test on human volunteers and because of this, they were quite limited to what kind of crash tests they could actually do. Now, this led to the development of crash test dummies and nowadays, almost all tests on animals and humans are replaced with the crash test dummy tests. Now, when you think of it systematically, this is quite similar to what we have in medicine. So, we test on animals first in the clinical trials, then we progress on testing in human volunteers. So, that makes you wonder where are our crash test dummies? So, the question here, if you think of it systematically is what would we need to build something like that? So, a crash test dummy is a physical representation of a human body. It is designed to predict the body’s movement during a simulated vehicle impact. Now, in medicine, what we would need to create a crash test dummy is a logical representation of a human body that actually predicts the physiological response to a certain treatment. Now, the question is not so much “Should we build something like this?” it’s more “Is it possible to build something like that?” and I think the answer to it is yes. We can create what we call an avatar. Now, an avatar is basically a digital twin of a patient. It is a physiological representation of certain aspects of a human being.

Dr. Frank Maddux: Tell us a little bit about how you create the avatar in the physiology that you worked on related to anemia and the production of red blood cells.

Doris Fuertinger: Red blood cells are absolutely fascinating. They are a product of the highly regulated and complex process called erythropoiesis. So, it starts in the bone marrow and then it progresses through several developmental stages and the cells enter the bloodstream, where they circulate for around four months. Now, the hormone called EPO is the driver that facilitates the red blood cell reproduction and is actually produced in the kidneys. Now, to have a normal level of red blood cells, a human being actually needs to produce a vast amount of new red blood cells every day. So, if you cut your finger and you bleed, you don’t see any red blood cells because they are too small to see with the naked eye. However, if you would line up the red blood cells side by side that a healthy man needs to produce every day, this would amount to about two kilometers, which is five Empire State Buildings every day. So, this is just mind-boggling. Now, if the red blood reproduction and the red blood cell loss does not coincide, so, the red blood cell reproduction cannot keep up with the loss, a condition called anemia happens, so, basically, the lack of red blood cells and anemia can have many causes, but kidney disease patients are especially at risk to develop anemia because we have here a perfect storm. We have a deficiency of EPO, this hormone produced in the kidneys, and also the red blood cell lifespan is about half in kidney disease patients. Now, what we do in the clinics is we can provide, we can administer EPO, this hormone to our patient population. However, only at discreet time points and this is a far cry from the very delicate regulation that the human body does by a minute by minute basis. So, inevitably, what happens is that the red blood cell levels tend to fluctuate or we cannot reach targets. Now, for a human being, it’s really difficult to grasp how even for healthcare professionals, how one single specific dose of EPO actually influences this red blood cell reproduction of a specific patient. However, despite what you might think when you look back at your high school math education, mathematics is not there to make simple things complicated, but actually to make complicated things simple. So, when we go back to what we know about anemia so that the cells are reproduced in the bone marrow, but they have to undergo several developmental stages, all of these things can take this knowledge and express the physiology in mathematical terms and this is what my group did. So, we actually created a very comprehensive mathematical model that’s based on physiology and we can use this model now to predict how red blood cells are reproduced and quantify how an EPO dose influences this. Now, obviously, there are some differences between patients. Now, when we go back to the crash test dummy analogy, although, for example, Dr. Maddux, you and I are quite different in many aspects-- gender, height, weight, age, and so on, and we might react to different stimuli differently, there are certain things that are still the same. Similar as in the crash test dummy example, both of our bodies have to follow Newton’s Law of Inertia in a car crash and both of our bodies follow the same basic rules how we reproduce red blood cells. Now, the trick lies-- how do you get the details of one specific patient into the model and this is where we use clinical data from individual patients. We use the data from the patient together with the mathematical model to actually understand what is the specific red blood cell lifespan of this patient? What is, for example, the bone marrow size, which is also quite different in patients? Then we have created something that we call Emmas or Johns, anemia avatars and these avatars that are specific to a certain patient enable us to actually predict how these patients would react to a certain dose of people.

Dr. Frank Maddux: It’s fascinating. I have two follow-up questions to that. One is to what degree do you find that environmental factors have to be incorporated into the model, like if the patient misses a treatment or a treatment gets delayed or they have an infection or they’re hospitalized, how do you account for those things? Number two, I’d ask for you to comment on what degree of computing power do you need to try to solve all of these mathematical equations for the patient.

Doris Fuertinger: This is actually-- to answer the first question, it’s a terrific question. So, it’s not only about the patient and the patient avatar. Patients do not live in a vacuum. They are not living in a research lab. So, as you said, they sometimes miss treatments and other things can happen. So, they live in an ecosystem and this ecosystem very much determines or contributes to how they actually react to treatment or how we can treat them best. Now, what we did is we created also a virtual clinical environment. So, the idea is that our avatars actually are being hospitalized from time to time. They sometimes decide they do not want to show up for the treatment. They sometimes do not take the pills or their attending physician overrides the recommended dose and this is virtual clinical environment actually helps us to make our predictions way more realistic. So, to kind of heed this together or put the puzzle pieces together, the virtual clinic consists of several things. So, we have the patient avatar, which is basically a physiological model that is informed by the specific patient data and then a cohort of avatars is being put in a virtual clinic environment, so, in a virtual ecosystem and all of this together forms a virtual clinic, so, for example, a virtual anemia clinic and this enabled us to actually predict or run a virtual anemia trial that has very high predictive power.

Dr. Frank Maddux: Let me ask you another question and that is you built these avatars, you’ve collected them together in a cohort, in a virtual clinic, and you begin to run a clinical trial on them. I’m assuming all of this is being done in silico, inside the computer, and just describe a little bit about what computing power you need to have to produce the avatar, how efficiently you can do that, and how do you then run the clinical trial.

Doris Fuertinger: Depending on the size of the avatar cohort and what kind of or number of clinical trials we want to perform, you might not be able to do that on a normal-- well, you’re not able to do that on a normal laptop and you might not be able to do that on a more powerful standard PC. But you actually have to go to something like cluster computing. Why? Just to give you an idea of what kind of computations are involved-- so, when I talked about clinical trials, I mentioned that you basically are restricted to test one or two different treatment options at a time. So, for our virtual anemia clinical trials, we actually tested 60 different treatment protocols and we used the virtual avatar cohort that was the size or larger, actually, than the size of the three largest anemia studies in hemodialysis patients combined. Now, during the simulations, we created 400,000 years of-- anemia patient years of anemia treatment. So, this is no small feat and this is something that would have been absolutely impossible in a clinical trial and to achieve something like that, you need quite high computing power.

Dr. Frank Maddux: So, it’s very interesting to me that we’ve been able to refine our actual clinical algorithms in practice by iterating different adjustments to the program and testing them through the virtual clinical trial methodology you described. Tell me a little bit about how this has been received in the community, the academic community that you work with and I think this has been groundbreaking work that you all have done. I’m curious how it’s been recognized amongst mathematicians and other like-minded individuals that you work with.

Doris Fuertinger: Virtual clinical trials have been proposed as a means to enhance the repertoire of clinical trials that we have and the idea is that when you build realistic simulations of a clinical trial that this can actually help to test safety and limitations of treatment protocols. This is something that is acknowledged or is starting to get acknowledged by the scientific community, by research and development departments in industries, by academia, and also, by regulatory agencies and actually, so, my team and I are spearheading this development of new and advanced techniques to actually incorporate not only real life patient data, but also clinical operational data. So, this has piqued quite some interest in the scientific community and also, among our peers.

Dr. Frank Maddux: So, how many avatars have you built in aggregate today? How many total avatars has your team put together?

Doris Fuertinger: For anemia, we have more than 7,000 anemia avatars readily available. We have actually since-- so, we have done already several rollouts of virtual clinical trial developed anemia protocols. So, basically, when we think back to the difference between clinical trials and virtual trials, there is this notion that to some extent, we can enhance what we see in clinical trials because we can test on patients or include patients that we usually do not include in clinical studies, like elderly and frail patients, patients with a lot of comorbidities, patients from all racial and economic backgrounds, which is impossible in a clinical study. So, having done that, you would think that the virtual clinical trial gives you an additional arrow in your quiver, so to say, to actually advance patient care. One thing we should always keep in mind, although it is a great tool, whatever you test in the virtual clinical trial, so, only on the computer, at some point needs to be confirmed in a clinical setting. So, we were really lucky that we had the support and help from Dr. Peter Kotanko, the Research Director of the Renal Research Institute, and Dr. Jeff Hymes, the Chief Medical Officer of Fresenius Kidney Care Clinics, that actually helped us to roll out our first anemia algorithm in 2013 and since then, we have come a long way. We have repeated this process for another drug used for anemia treatment, we have advanced our techniques that we have to actually do that, and there’s another development, a recent development, where we try to drive this to the next level, a fully personalized anemia care. So, the idea is that every patient in a dialysis facility gets his or her own anemia avatar and before a dose is administered to the patient, it is actually tested on the avatar and only if we see great results, it will be recommended to the clinics that will provide this EPO dose to the patient. Now, we are currently testing this concept in the clinical study in the Renal Research Institute clinics and I’m quite proud to say that our first participant very recently has completed the six-month study period and we really saw terrific results. So, the patient came from previously high fluctuating hemoglobin levels and was completely stabilized within the target for the entire study period. So, this was a great achievement for the team.

Dr. Frank Maddux: That’s a great advancement in trying to get towards more precise prescribing for patients with kidney disease and anemia. We know there are new drugs coming into the anemia space, the hypoxia-inducible factor inhibitors, and I’m interested in whether you’re working to model how those drugs might impact the anemia care that patients receive.

Doris Fuertinger: Yes. So, we are in the middle of the process to develop a physiological model for these new HIF stabilizers and we are about to complete the first version of this model and start to build this virtual environment around it, which is really exciting for us as a team, not only because it’s a new challenge, also because here, it’s a new drug where nobody has data on, despite what we know from clinical studies so far. So, when we look at-- so, currently in medicine, most of the things people are doing is based on machine learning and while machine learning algorithms do work, one of the prerequisites is that you have a huge amount of data readily available and this is not available for these new types of drugs and it will not be available for months or years to come. So, here we are really spearheading and driving methods to actually go beyond what we see from machine learning to help better the treatment of our patient population.

Dr. Frank Maddux: What are some of the other areas of physiology that you think are ripe for being able to be modeled mathematically and to develop avatars for other physiologic systems? Anything come to mind in particular?

Doris Fuertinger: There’s, on one hand, a lot of things that are difficult with kidney disease patients. So, beyond anemia, bone disorders and vascular calcification is a main driver of morbidity and mortality in this patient population. So, in that area, we also have developed and have a model available that can be used to run virtual clinical trials and we actually-- so, there is a condition where PTH, this is a hormone that is produced by tiny glands in the neck, and when this is elevated, your bone basically gets absorbed. So, it gets eaten up and there are a number of drugs out there that can help to manage this, but they are quite expensive and for example, Cinacalcet, a drug that’s very frequently used to treat this condition, is usually given daily. However, because of a high pill burden and stomach side effects, patients do not always take them. So, what we investigated in a virtual clinical trial is if instead of giving the patient the opportunity to take them daily, maybe we can provide this treatment three times a week in the dialysis facility and what we saw is that actually, this three times a week administration, for a lot of patients will result in the same kind of control of these PTH levels than the quote-unquote “daily administration” at home. So, because of the results of this virtual clinical trial, thousands of patients in the Fresenius Kidney Care Clinics have switched to this three times a week administration scheme. Now, this is an example in the kidney disease realm, I say, but the virtual clinic concept is not a one-trick pony or two-trick pony. In principle, the virtual clinic concept can always be applied when science has, medicine has elucidated the disease process well enough so that we can actually describe them mathematically and although this is not true for some diseases like everything that’s connected to mental diseases, this is true for a lot of the great and agonizing maladies of humankind. So, for example, and I’m sure every one of you knows a friend or relative or colleague that suffers from osteoporosis, a condition that fractures mostly the bones of women or heart disease and diabetes, which is a driver of morbidity and mortality globally and all of these diseases and more are actually available for this virtual clinic concept and we think that using these ideas that we have developed and are establishing right now, this can lead to a situation where we can actually-- we’ve reduced costs way faster than we currently do, improve the treatment and patient care of this patient population.

Dr. Frank Maddux: Doris, thank you so much for being with me. Anything else you’d like to add that maybe we haven’t covered?

Doris Fuertinger: One last thing-- as I said, the virtual clinic concept is not a one or two-trick pony and is not limited to the area of kidney disease and we actually have-- team has worked on a proof of concept in the area of osteoporosis and here, I really need to mention Dr. David York from my team, who was driving these efforts and what is very, very interesting is that we see that for osteoporosis, where there are a lot of different medications out that can be used to treat it, but there’s no really good guidelines how to actually combine them and this field can actually make great use of the virtual clinic concept because in these simulated clinical trials, it’s quite easy to test different combinations of drugs and see what is the best option for the patient and this is actually what we have done and it’s very exciting that we see that the-- if you take three drugs, for example, and you just change the way-- the order in which they are administered, then you can actually influence the short and the long-term outcomes for the patients significantly and this is so exciting and something that we would have not been able to see in clinical trials for a long, long time.

Dr. Frank Maddux: Dr. Doris Fuertinger has been with me today and I just want to congratulate you and your team on the work that you’ve done over the years and the work that you have to do in these many areas of physiology. I think it’s been quite groundbreaking and thanks for joining us.

Doris Fuertinger: Thank you so much for having me.