Insights from the OptumLabs Research & Translation Forum.
Evidence in context
OptumLabs partners are putting evidence that matters to patients at the center of decision-making. It starts by talking with patients, and leveraging real-world data, and more.
At Mayo Clinic, Dr. Victor Montori believes we can't wait for the right evidence to trickle down and improve health. He urges clinicians to understand both the biology and the biography of each patient. This can transform "evidence-based care" to "evidence-based care."
A patient revolution for careful and kind care
Dr. Montori's focus on shared decision-making at Mayo Clinic identifies opportunities to open "the black box" of health care. This allows patients to engage in treatment decisions that align with their lives and values.
Speaker: Victor Montori, MD, MSc, endocrinologist and professor of medicine, Mayo Clinic
- We're very fortunate to have Victor Montori join us today. I have to confess, that we've tried to get Dr. Montori to attend our RT Forum for five years, and this is the first time we've been lucky enough to have him, and to have him speak to us. So, thank you. He is a Professor of Medicine at Mayo Clinic, who has done a lot of important work as a practicing Endocrinologist and expert in evidence-based medicine, and shared decision-making. Dr. Montori developed the concept of minimally disruptive medicine, and works to advance person-centered care for patients with diabetes, and other chronic conditions. He runs a shared decision-making lab at Mayo Clinic, and runs multi-million dollar trials on the use of shared decision-making tools, to look at different ways that patients and physicians can have conversations to determine what works best to provide care that is consistent with patient preferences, and improve their comfort level and satisfaction. And he recently published a book titled Why We Revolt: A Patient Revolution for Careful and Kind Care. I'm sure available on Amazon.com, for those who are interested. Today, he's going to share some of his perspectives on this topic. Dr. Montori, I look forward to your talk.
- Thank you. We've been working on research for the last almost 20 years. From the beginning, we decided to take no funding from any for-profit corporation, so I have no financial disclosures from a research perspective. And as Paul mentioned, I co-founded a organization called The Patient Revolution earlier last year, and then, published last year this book Why We Revolt. All the proceeds from the book go to that organization, I don't profit from it at all. Some of the ideas in the book will be discussed in my presentation today. One of the thing, so, the main theme of evidence-based care these days, can be summarized in these taglines that belong to organizations, very much like OptumLabs that may have as their tagline, better evidence for better decisions. Or they might be focused as we will see later in the program, they might be focused on the challenge of translating the evidence that we have into practice. Part of my points this morning is to point out that this is a top down approach. One that expects evidence to trickle down and to then impact health. So, if we just get the evidence right, people will be healthier. It's almost like a form of trickle down EBM. If you think about the effects that this has, this is one of many grafts that have been put forward to describe the evidence ecosystem. In the evidence ecosystem, or the learning healthcare system or whatever other model that you might know, it's portrayed as a machine that is able to produce knowledge and then push that knowledge down into the, into care. Making healthcare, the healthcare delivery a black box in which you bring in the right input and you expect a right output. And if the outputs are not right, which you can capture with your evidence machine, you just changed the inputs for policy and you expect the outputs to come out the other way to be right. My second point in this presentation is to draw attention to that black box. That if you just get the evidence right, there's no real reason you should expect the outcomes to be right unless you pay attention to what happens, to how this information is processed into that black box. So, again, the assumption that I think many people in the room make, superficially, I suspect, not after you think about it, is that if you produce good evidence, outcomes for people like Maria Luisa will be better. So, let's understand what this big evidence-based care looks like. So, here is Maria Luisa. Maria Luisa is a Peruvian like I am. But, she lives in Anchorage, Alaska with her son and her two granddaughters. What do you see in this picture? Just yell it out, again, remember, hearing loss. So, you have to yell it out. What do you see? Look at her eyes, look at her expression. Don't look at the big mole on her cheek. That's what all dermatologists, "She has a mole on her cheek, what is that?" No, no, that's not the point, that's not the point.
- Isolation, what else?
- Loss, fatigue, keep going.
- She's got a wearable.
- A what?
- A wearable.
- Yeah. Some people say, "Oh, that's a old-fashioned Fitbit." No, no, no, that's one of those things when you fall, it alerts people that you fell. It's not an old-fashioned Fitbit, or maybe it is an old-fashioned Fitbit, yes. What else?
- Overwhelmed. Yeah, lost, depressed. None of you are yelling, "Healthy!" Even though she might--
- She appears to be in an institution.
- Because of the fire extinguisher, isn't it? No, no, that's actually her home. They might not trust her with fire. But, that's actually her home. So, she is living independently. So, for some systems, that'll be a claim of health. Yet, none of you are looking at her will claim that she is healthy. In fact, she doesn't appear to be. The way we would be thinking about her is how might we respond to her challenge. You know she is all the things you said. She has obesity. She likely has multi-morbidity. She's frail, she has limited life expectancy. And in that translation of evidence into practice, we would like to respond to this situation. That begins to show us a different set of problems. First, as people accumulate multiple chronic conditions, they become more difficult to categorize. So, you have your roving ideas thing and you have Alzheimer's. Yeah, yeah, but the person with Alzheimer's also has diabetes and has a bad hip and has osteoporosis and is depressed. And you have another initiative on opioid addiction. Yeah, yeah, yeah, but that person also has other biologic, so the problem is when you pick up any of these labels, you're picking up a whole other range of labels that make it very hard to continue to articulate that Maria Luisa is a person with diabetes or is a person like she is, a person that lives with chronic renal impairment and gets dialysis three times a week. What disease describes her best becomes an exercise in futility because she's all of those things. And as we respond to Maria Luisa, because the other thing I didn't show you in the initial picture is that Maria Luisa has plenty of healthcare. So, she is not only looking unhealthy, but she's looking unhealthy despite having absolute complete total access to very aggressive forms of care. You can see all the poly pharmacy in front of her. We would like our evidence to be influencing what Maria Luisa gets. But if you look carefully and you say, "Well, if Maria Luisa only were to get evidence-based care," so maybe if there's a randomized trial, that will be the tablet that she'll be taking. The problem is that that other tablet, also supported by randomized trial, and that one and that one and that one and that one and that one. Now, of course, when you get to the things that she's eating and drinking, well, that's observational data. We don't have randomized trials there. And then the other bit is this. As people accumulate multiple chronic conditions, we get into troubles with our guidelines. Our guidelines tell people what to do in terms of treatments to give and so forth. And patients with multiple chronic conditions will have guidelines that eventually will conflict. And as a result we may have drug-drug interactions or drug-disease interactions, mostly with renal impairment. Which might make the case that if you adhere to the guidelines very strictly, you might cause harm. But, that's not the main problem of dealing with the problem with Maria Luisa through evidence or with evidence. It's that most of the response would be biological. And the situation of Maria Luisa is not just biological. In fact, when you were all yelling your answers of what you saw, what you described was biographical. Material deprivation, isolation and loneliness, being stuck at a dead end. That's what you were, overwhelmed, depressed. So, in responding to Maria Luisa's situation, the clinician at the front line has to keep their gaze oscillating between the biography and the biology so that they can respond coherently to the situation of Maria Luisa. So, our evidence that focuses exclusively on the biology of things, as it pushes down, it has a blind spot for the biography and integration of that biography into care. So, we delegate that integration. We delegate that translation to our front line clinicians. But, those front line conditions are in trouble. Why are they in trouble? Well, first of all, we have the severe problem with burnout. Callousness, lack of empathy, inability of clinicians to connect emotionally, intellectually with the person in front of them. Second, we have a problem where the patients are being perceived as a blur. They're seen as a blood test or as a biopsy result, as a statistic in population health or simply just too fast because we have little time in the appointment to pay attention to them. And whatever time we have is dedicated to an agenda that is bloated. An agenda that half of it, at least, it's time is dedicated to feeding a computer system. The other half, we turn to the patient to ask the questions the computer system is demanding from us. So, to compensate for the fact that the patient is a blur, we have tools that we use in our encounters, guidelines we use, targets for disease control. We use shared decision-making and treatment intensification to respond to the person we think we have in front of us, response that is for patients like Maria Luisa. But, it's not for Maria Luisa. And for the clinician, that is not the job. The job is not to respond for people like the person in front of me. The job is to respond to the situation of the person in front of me. And that job is made incredibly difficult. So, the end result of these evidence-based guidelines and care protocols and quality measures and specialist care that is actually disease-oriented and context-blind, is that their regiments are increasingly complex. There is very little prioritization, there's almost no coordination. If you look at care coordination programs, most of them are focused on achieving the metrics rather than actually making healthcare fit in the lives of people. The result is the face that you see in Maria Luisa and in her caregiver, which is the face of being overwhelmed. Now, I want to bring to you a little bit more insight into what is this work of being a patient. When I say that are overwhelmed, clearly I'm pointing out that people are having to do more than they can do. People have estimated that if you have three chronic conditions, you're consuming six medications per day, you're having six clinic visits a month, 60 hours a month of health-related activity. When you increase the number of conditions, that work doesn't grow exponentially, it goes up to some extent. And you can see that it amounts to what could be a part-time job. If you're more visually oriented, this is the same date but in visual. That blue part is the work of getting the medicines set up, refilled, available, organized and taken. The yellow stuff is self-care, diet, exercise, unsupervised therapies. Red is supervised therapies and green are clinic appointments. So, you can see, as diseases accumulate from the top to the bottom, the amount of time that people have to put into healthcare goes up. But an interesting picture emerges when you ask patients to take pictures of the work of being a patient. So, this is a project that we did with diabetes patients. And these are the pictures that they send to us. So, you can imagine, diabetes patients, one of the projects they have to do everyday is to look at food labels and nutrition labels and understand them. They have to learn how to use these devices to self-monitor. Although, this picture is not to show us the difficulties with self-monitoring of blood sugar, this picture was given to us by the patient to show us that this was, I think, the fourth or fifth glucose meter that they had received. As their insurance was changing providers of glucose meters, they were getting a new device. Each device has a different and separate interface that they have to learn. And, of course, uses different glucometer strips. So, the patient had to throw away all those strips, get new strips, and learn the new device interface. The effect for me, I'm a diabetes doctor, when they come to see me in the clinic is that we download this data so we can study together and, oftentimes, the meter's completely empty. The patient's dismayed, "I've been checking my sugars "four times a day. "How can it be empty?" And then we look carefully. The date on the meter is 1982. And, of course, when you connect it to the computers, it downloads the last three months, there's nothing in the last three months because all the data is in 1982. They just never set it up to have the time and date correctly because the interfaces are difficult and complex. Many patients show us this picture. Wait. Wait, looking at the wall or in the waiting room. And just sit there and wait. And then you hear the steps, that's your doctor coming in. Oh, no, they went to next door. And often the door opens up, "Oh, sorry, wrong room," and then they walk away. And then you wait and you wait some more. This is not a hospital area, this is somebody's home. They set up a cupboard for their insulin infusions, pump supplies. So, it takes work. This is perhaps, to me, the most interesting one. When you look at the diaries of patients that are trying to make everything work out, you see how they have to weave in the work of being a patient, with the work of life. And you see their notes about, "I have to take my kid to the speech therapist "and I have to go to work, "and I have to see the nurse, and I have to go somewhere," and life and healthcare become one. And those who are skilled enough to make all of this work, and of course, the people with most comorbidities, with multi-morbidities, are the people with the least resources, the lowest social economic style, the least education, the least ready to engage in this incredible feed of self-management and self-care. Such, 75% of our patients, we've begun to survey patients with a burden of treatment questionnaires. 75% of our patients are reporting high levels of treatment burden, and some of them declaring them unsustainable. But a lot of this work is the result of translating this evidence into practice. And so one of the concerns that we've had is that perhaps the evidence that we have to translate into practice is actually giving us a sense that the treatments that we can offer people are better than what they really are. And that gives us a sense of, you know, down the torpedoes, let's just push for these treatments to make sure they can get to patients. This is a very, I think, helpful graph that shows the evolution of antidepressant efficacy therapy trials and the evidence of it into practice where you can see the effects of reporting bias, then publication bias, and then citation bias on how what appears to be a 50/50 proposition in terms of efficacy on antidepressants at the end of the days communicated as if antidepressants were universally effective. As a result, we need opportunities to develop evidence, at least at the surface, that is independently produced of those who are going to profit from its results. That is error-proof with increasingly improved methods, which is, I presume, some of the concerns of people in this room. That is presented spin-free which is a concern when organizations have to demonstrate to their stakeholders and to the public that their work is actually worth funding. So, we hype it up. And the problem with hyping it up is that people will believe that what we're doing is really, really important. Then they want to push it further and more strongly to the point of care to people like Maria Luisa potentially overwhelming their capacity to self-care. It has to be, of course, fully reported so that people have a sense, not only of the good news, but also of the caveats and the potential balance of good and bad news that the full report may show. Skepticism has now given way to hype and hype gets translated into over-enthusiastic guidelines and over-enthusiastic guidelines into quality measures and quality measures into a reduction in the range of things that we can do at the bedside for patients and for patients, an increasing amount of work. There's been proposals to bring the voice of the front line of the clinicians who are aware of the work of being a patient and of patients who are aware of living with chronic conditions so that they can set out priorities of work. So, Picor and other organizations have put a big effort in trying to bubble up concerns of the front line. So that this top down approach of evidence into practice doesn't lead to problems like Maria Luisa's. You can see that people have looked, about three percent of clinical trials are responding to patient priorities. The rest are responding to industrial priorities. I have a drug, I have a device, I need to get it into the practice so I'm gonna do a study, get it approved, and then I'm gonna push it down with my results. I'm gonna trickle it down so it will improve health. So, then we do studies not because people at the front line need to know the answer, but because we need to do it at the top of the pyramid. We need to know the answer so that we can have a successful product launch, a successful industry. The best example, probably the worst example I can give you, but, to me, it's a fun example, is the Apple Watch that can detect this heart rhythm stuff. Oh, I wonder what this is good for? Let's do a big trial, right? I mean, I don't know anybody walking down the street here in Boston, asking themselves. "Hmm, I have this problem. "I'm sure the Apple Watch is the solution "to that health problem." It's not a bubbling up of the problem, it's a trickle down effect. If you look at what makes healthcare evidence useful, it's not what was being produced massively. But, perhaps it's something you're correcting. Perhaps this is something you are making better. It has to be responsive to the problems of the front line. It has to be responsive to the problems of Maria Luisa and of her clinicians. So, we need to have a deep understanding into what are those problems. We need to be inside that black box and understand this problem of being a patient, the work of being a patient, and how the next therapy, the next obligation, the next must do falls into the lap of those clinicians who are overwhelmed and tired, et cetera, and of the lives of patients. It has to be directly applicable to the situation of Maria Luisa and her biology and her biography. Again, we need to then develop a deep understanding of what those means and then have studies that are very inclusive of people, regardless sometimes of the limitations in their biographical aspects, which makes them terrible research participants, but necessary research participants if we want to develop evidence that is directly helpful. Your work might actually overcome some of those challenges. And it has to be practical. It has to be something that people can make fit into complicated lives. So, it's the other way around, isn't it? It's not that you will develop evidence and then push it down, make it trickle down into the lives of people like Maria Luisa. Keeping that space between, as a black box, but it's the idea of understanding deeply what's going on. What are the problems of health and healthcare at the front lines? And then developing the evidence that will respond to that problem. That inversion of the arrow means that you cannot be doing this research just by talking to policymakers and staying behind your computers looking at databases because both of you, your stakeholders and yourself are looking at the problem from outside the black box. And without the relevant partners at the front line, engaged throughout the process, you will miss this opportunity and you'll be writing a lot of papers about the barriers to the adoption of your insides. So, at the end of the day, this useful approach must be participatory. You need to have a bunch of patients and clinicians in this room, not just their leaders. But the people at the front lines that are experiencing the problems, it needs to be collaborative. It needs to be developed in context. And it needs to be multi-methods because you cannot, just with quantitative methods, gain a full understanding of the narrative of the lives of the people that you're trying to affect. Now, there's another problem and I would like to finish with this. If you do this right, you would expect that eventually there will be an opportunity for patients and clinicians to come together and care will happen. But, I told you, the healthcare system as it is right now does not enable us clinicians to see the patient in high definition. So, what we need to do is we need to reform the healthcare system so that it can actually care with evidence, with the evidence that you produce. And to care with evidence, we need to create an opportunity for unhurried consultations. Every healthcare organization is focused on access and making sure we can see everyone and we can do it efficiently and effectively and we're missing the point. The point is we need an unhurried conversation to care for people so we can see them in high definition. Maintain this gaze oscillation between biology and biography so we can get the answer of what care this patient needs right. And we need to identify a sensible solution. A solution that, at the end of the day, makes intellectual sense. It's based on the best signs, makes emotional sense, feels like the right thing to do, and makes practical sense. People can weave it into those very busy lives, the very complex and complicated lives that they live. It's only when we have all these elements in place that patients will feel that they are receiving care, that they're being cared for and about and clinicians will recognize that their calling, that the reason they went into healthcare is being honored by the system in which they work. And when that happens, they won't provide care, they won't deliver care. They will recognize once again that care is, in itself, a verb. They will be able to care. They will care. That is what we need to be focused on. So, when we take that evidence for antidepressants that I showed you, one of the efforts that our group has done is to develop tools that actually corrects for some of the distortions in the evidence and bring those tools to the point of care. So, here's a shared decision-making tool. It doesn't appear like that because it will be overwhelming. It builds up over time in an electronic fashion. And it's used, not by the patient alone at home, which will further their isolation and delegate the intellectual work to them, but rather it's used in the interaction, in the clinical encounter, forcing an unhurried consultation, a discussion about the issues that matter to this patient so that we don't care for people with depression. But, we care for this person with depression. The point of the shared decision-making activity is not to translate evidence into practice, furthering the trickle down effect, it's actually to bubble up the challenges that this patient's having and seeing how we can respond to those challenges with evidence, evidence that you dedicate time, effort, and resources to build, hopefully, in an unbiased and applicable fashion. To make all this work, we need to change the healthcare system and create a way forward that will make this the easiest way forward. That means that we cannot have a healthcare system based on greed. We need a healthcare system based on solidarity in which this idea of care doesn't become an exceptional oasis in our society. But, it's a demonstration of our commitment, common commitment of caring for ad about each other. Think about an natural disaster. Any movie, any video you see of a natural disaster, there'll be a stranger trapped in harrowing water and a string of strangers locking arms, eventually reaching this old man in a van and pulling him out of the water, saving his life. Nobody in that group, in that line, knows each other. So, there's something deep in their DNA that is telling them that the human thing to do is to care for and about each other. What's strange is that we wait for tremendously limit situations, tragedies, disasters, to demonstrate that to ourselves. We can do this routinely. We can care for and about each other. And healthcare should be the place where we can demonstrate that. Relations and incentives means that we don't trust anyone in healthcare. We train people for four, six, eight years, we put them in the front line and then we don't trust them with their judgment. We don't think they're gonna do well. Instead of developing a gravity in the system based on integrity, we set up rules and regulations that make the system rigid. And as a result, when people come in for help, make the system cruel, which is incredibly problematic. We're focused on efficiency but that's the wrong idea. We don't want waste. But there's something different than efficiency that we should pursue in healthcare's elegance. Think about a gymnast or a ballet dancer. Don't worry, I'm not going to... There is no wasted movement. Everything is in place. There is no waste. But there's a tempo to what they do. There is no haste. There is no rush. It all occurs when it should occur. That's elegant. And we need a healthcare system that learns to be elegant, that gives up this idea of efficiency, which is great for productivity, for production. But it requires that we make clinicians and patients interchangeable. If you do that, for chronic condition, you won't develop the relationships that when a disappointment occurs, when a bad outcome occurs, those are the relationships we fall back. That's where our resilience comes from. If you make everyone interchangeable, don't be surprised that your clinicians will burn out and your patients will look like Maria Luisa. So, we need elegance, not efficiency. We need to take care of people, not people like this. But for this person. And we need to create systems that allow us to do that and that moves us away from transactions to relationships of love. Because when we care for and about each other, we make ourselves slightly vulnerable. And getting ourselves closer to each other, we are able to respond to the challenges of the patient better with empathy, with love. It's interesting because when I mention love in healthcare, people go like, "Uh, that's a little icky, love." Somehow when I mention profit, everything is fine. I don't understand that. The vision, then, is to move from evidence-based care to evidence-based care. Where evidence-based is an adjective to what we're trying to accomplish. What we're trying to accomplish is not move evidence into practice. What we are trying to do is care. And we want to care with the best possible evidence, with the best possible science. That's the role that all of you, I think, should be playing in this world. So, that Maria Luisa can improve. This is a picture of what happened after Ana, Maria Luisa's granddaughter, spent two years with us and went to visit her grandmother in Alaska. She noticed that she was a prisoner of the second floor. She was very afraid of going down the stairs, breaking a bone. So, she would stay on the second floor when her family left for the day. They put this little elevator and that helped her move around. That's Ana, there to your right. She also noticed that when she was on dialysis in the mornings, three times a week, the afternoon, she was completely wiped out. Maria Luisa was living a part-time life because three days of the week she was out. So, she wondered what happens if we move the dialysis to the afternoon. She got lucky. Not only did she have more time in the morning to do her crochet, which she enjoyed tremendously, but also two nurses in the afternoon spoke Spanish. She could only speak Spanish with people at home on the phone and with her family and that was it. She was very isolated, somebody mentioned that. So, here, two nurses that spoke Spanish opened up some of her world. But, what Ana did was absolutely genius. Was she connected with all of the dietary restrictions that Maria Luisa had, because of all her conditions, put them all and send them to a dietician in Peru and said, "Can you send me some recipes of Peruvian food "that we can cook here in Alaska for her?" So, she got those recipes back, found somebody that could come into the house and cook on Sundays and she could eat beautiful, delicious, wonderful Peruvian food for the rest of the week. If you know anything about Peruvians, that's the definition of life, right? So, she had access to this wonderful, wonderful Peruvian food. And this is the last picture I have of Maria Luisa. She got well enough, not only to feel well at home, but also to travel and travel back to her home country, travel back to Peru and have that enormous fish in front of her. I don't know if that was the healthiest option for her. But that is the picture of health, not the fish, her smile. Look, that is the picture of health. That is what we are trying to accomplish. It is that, that we're trying to do. We are trying to care so that people can live lives that are full of meaning and opportunity. We can only accomplish that, is my proposition to you, with careful and kind care based on the best available evidence. That last bit, that is your challenge. Thank you for your attention.
Combining different types of data creates more complete and reliable evidence. Dr. Peter Noseworthy and Dr. Xiaoxi Yao embody this concept by using real-world data to complement clinical trials at Mayo Clinic.
Personalizing care for patients with AFib
Learn how Dr. Noseworthy and Dr. Yao pair data from OptumLabs with a large clinical trial to personalize treatment decisions for patients with atrial fibrillation.
Speakers: Peter Noseworthy, MD, electrophysiologist and associate professor of medicine, and Xiaoxi Yao, PhD, MPH, health outcomes researcher-cardiology and assistant professor, health services research, Mayo Clinic
- So it's my honor to introduce Dr. Peter Noseworthy, Associate Professor at Mayo Clinic, Cardiac Electrophysiologist. I believe you trained down the street from here and Xiaoxi Yao, who'll be joining us, assistant professor with a lot of expertise in health systems research. Both of them together have done remarkable work related to shared decision making and in particular built a body of evidence around the appropriate use of medications related to blood clotting and many other areas. So, without further ado, Peter, we really appreciate you joining us.
- Thank you.
- So thank you very much for the opportunity to speak today. And Victor, thanks for a great talk. Every time you talk I'm inspired. Victor is a close collaborator of mine and a good friend. I've learned a ton from him, he's a great guy. The best guy to go get a beer or a Pisco Sour, but the worst guy to follow on a stage, and I'm feeling that, for sure, right now. Today, Xaoxi Yao and I are gonna talk about some of our recent work, and what I hope comes through in this talk is the idea of partnerships. So, I'm a frontline clinician, so the frontline clinician that Victor was talking about, and Xaoxi is a health outcomes researcher. And over the past several years, we've worked very closely together as a partnership to answer some of the questions that I see every day in my practice. And we're also gonna talk about the partnership between randomized trial data and observational data from data sources like OptumLabs. And we can see how these can be complimentary, and I think the power is in putting these things together, so putting the clinicians with the researchers, putting the trial data with the observational data, and trying to keep all the questions that we want to answer embedded in the clinical practice. So, I'm a Cardiac Electrophysiologist, I'm a Frontline Clinician, I see patients all the time, and my job is basically to figure out what the best treatment course is for an individual patient. And the clinicians in the room will know that that's a familiar process and it sounds like it should be relatively easy, but we're the recipients of the, sort of, trickle down EBM that Victor was talking about, and navigating that world can be very challenging. And, it's harder than I like to admit, especially since that's basically my job, but it's a very challenging process. This year, probably the most awaited and anticipated clinical trial in my field of cardiology was published, and you'd think that would make things a lot easier, but in fact, it has not. And that's because the results of the trial are sometimes hard to apply to an individual patient in the office, and even the interpretation of the trial can be challenging. And even though this was the main trial for us, it's generated an incredible amount of controversy and the blogosphere has gone crazy trying to interpret this study, and EP's and cardiologists are sort of lining up on one side or the other. But the frontline clinicians like me have to eventually apply what we've learned from these trials to practice. And, we anticipated that and a couple of years ago we've started working in this question. We started to try to recapitulate the clinical trial using OptumLabs data, and now that the trial is out at our data, we can put these together and make sense of the data. So first, a little bit about the clinical problem. I'm talking about treating atrial fibrillation. atrial fibrillation is basically a disruption of the normal heart action. So, instead of top chamber, bottom chamber, top chamber, bottom chamber in a coordinated way, the top chamber starts to quiver and the heart goes fast and it goes irregular. And as a result, patients feel a sensation of chest fluttering, they may be short of breath, they may feel tired and weak, and they come to us for help with these symptoms. Though, probably even more important is the fact that atrial fibrillation is associated with considerable morbidity and mortality. If you take a hundred patients diagnosed with atrial fibrillation and follow them for five years, some amount of them will have a stroke, 14 will develop heart failure, and almost half of them will have died within the first five years. So our treatment goals are twofold, we wanna improve the symptoms that patients come to us with, and we wanna try to prevent some of these long-term, adverse events, morbidity and mortality. And we have basically two types of treatment. We can treat with medications, or we can refer a patient for an invasive and costly procedure called a catheter ablation where they go under anesthesia, we put a catheter into the heart, cauterize areas, and try to restore and maintain sinus rhythm. And although it's relatively easy to take the medications, they are known to be less effective than ablation for rhythm control and for symptoms, and they do carry some side effects. ablation may be more effective, but it carries considerable costs, short-term risks, and the big question is whether it impacts long-term outcomes like cardiovascular mortality, heart failure, and stroke. And that's something that we didn't have good data on until just recently. So, what we're gonna do, if everybody has the app, we're gonna be able to vote on how we're gonna treat three specific patients. These are patients I could see any day in the office. And what I'm gonna do is we'll talk about these three patients, and then we will come back to the trial data, and then we'll go through some of the data from OptumLabs. And we'll see if we can figure out with better certainty about how we should treat these individual patients. So the first guy is William, he's 72, he's got hypertension, he has mild heart failure. His symptoms are shortness of breath and palpitations, and he's tried a medication called Sotalol and it hasn't helped, he wants help with his symptoms. The next patient is James, he's younger, he has absolutely no comorbidities, he's on no medications, he hasn't tried anything for his Atral Fibrillation, but he's in-and-out, he's being driven crazy by the palpitations, and he wants some relief. Third patient is Susan, she's 77, she has renal failure, she has the highest level of comorbidity of these three patients. When I ask her, do you have symptoms from atrial fibrillation, she's not certain. She feels tired, she wonders if maybe that's related to the a-fib, and she's tried our most powerful medication, Amiodarone, and it hasn't helped. So, for which of these patients would you recommend an invasive procedure to try to treat the atrial fibrillation, and you can vote for more than one patient. So, let's bring it up and see how this works. And we'll come back to this as we get more information. This is great, so, what I'm seeing is that most of you are thinking that William is probably the best candidate here, he's that sort of moderate-risk patient, followed by James, who is the low-risk patient, who's looking for symptom control. There's some hesitancy about Susan, who's the older patient with greater comorbidities and more vague symptoms. But let's come back to this as we look at more data. So we'll come back to the slides. This is the trial that I mentioned, it was called the cabana trial. And clinical trialists in the room will not be impressed by the size of this trial, but in our small field of cardiology, this was the largest and most important trial to date. It randomized patients to either medications, or catheter ablation, one-to-one, and it enrolled 2,200 patients. But it was the first trial in our field to look at hard cardiovascular endpoints. Mortality, stroke, major bleeding, or cardiac arrest. Now, the Achilles' Heel of this trial was that it took a long time to enroll. It enrolled over seven years, and ablation is already widespread used in routine clinical practice. So there was a lot of crossover in this trial, and it impaired the ability for us to really find out what the attributable risk reduction is with ablation. So, in the drug-treated arm, 30% of people crossed over to catheter ablation, and 10% of the patients randomized to drugs never got the ablation at all, usually due to logistic reasons. Now, this is common for all kinds of surgical or interventional trials, but it makes it hard to interpret these data. What did it show? Well, the bottom-line result in primary analysis of this trial was that it was a neutral trial. Ablation failed to reduce hard cardiovascular outcomes. You can see the Kaplan-Meier curves here. However, in pre-specified analyses, and this is where the controversy comes in, in per-protocol analysis there seemed to be a difference. So, everybody in EP either lines up on one side or the other. Do we totally lose the effect of randomization, and this is garbage, or are we actually finally understanding the impact of ablation and rhythm control on outcomes? And as well, in a treatment-received analysis, the effect was even a little bit stronger. So, a lot of controversy. Clearly we need more data, and we had anticipated this a couple years ago. We thought the question was going to be one of generalizability and external validity of the trial. We didn't anticipate the controversial primary findings. So, Xiaoxi and I put through an R21 NHLBI grant, and we tried to recapitulate the clinical trial using observational data. And this is what we see for the generalizability piece. So, we identified 180,000 patients in routine clinical practice who were treated with drugs, or with ablation. And, applying the cabana inclusion and exclusion criteria, what we see is that a minority of patients looked like James, they're low-risk. In order to get into the trial, you needed to be 65, and you needed some cardiovascular risk factors. About 4% of patients with atrial fibrillation in practice looked like that low-risk profile. So the trial does not apply well to, I hate to use the term, but patients like James. Also, about a quarter of patients would have been excluded because they're too high-risk, like Susan. Whereas about three quarters of the patients are more like William, and could have been enrolled. So you know a little bit now about the bottom-line result of the trial, and you know in broad strokes whether these patients would or would not have met inclusion criteria. Let's vote again and see what people think. Has this been helpful? So, what I am seeing here is that fewer of you want to refer Susan for ablation, and I think that's valid, because we don't anticipate strong morbidity, mortality benefits of ablation, and she probably is at the high-end of the risk spectrum. Maybe fewer of you are uncertain. But I wanna hand it over now to Dr. Yao, who's going to talk about how we use OptumLabs to try to supplement what we've learned here from this controversial trial.
- So now, I hope, many of you feel confused. Because there's some point here, right? Because, from the clinical trial, we actually don't know whether ablation has a true cardiovascular benefit. This trial took about 10 years to complete, and cost many millions of dollars. But at the end of the day, we don't even know what this trial found! So, now I'm going to use the OptumLabs, they had to see whether we can make things clear. Oh, is that? Okay. So, we divided our patient population into two groups. Those treated with ablation, and those treated with drugs. And then we use propensity score weighting to balance them on 90 baseline characteristics. So after propensity score weighting, patients in the ablation cohorts, and patients in the drug cohorts look almost identical on every baseline characteristic. And then, we compare the two treatments for the primary trial endpoint. And, yeah, our first analysis, we limited to patients who are eligible for the trial. We want to replicate the trial analysis in a group of patients who are eligible for the trial. So, in this analysis, we found a hazard ratio of 0.7, which is very similar to the trial per-protocol analysis. This is not very surprising, because, remember, the per-protocol analysis excluded patients who deviated from the protocol. In other words, the per-protocol analysis excluded patients who cross over between the two treatment arms. So, in our OptumLabs analysis, we actually didn't have any crossover, because we just divide our patients based on the treatment they received, right? So we have no crossover, and we found the hazard ratio very similar to the per-protocol analysis. But our results was more significant, because the sample size was larger. We also did a pseudo-intend-to-treat analysis. We want to simulate the trial intend-to-treat analysis. In the trial intend-to-treat analysis, there was crossover between the two treatment arms. But, in our analysis, we didn't have crossover, but we want to create it. So how we did that, we just randomly select some ablation patients and mix them with the drug-treated patients. And then we randomly select some drug-treated patients, and mix them with the ablation patients. So we just created this crossover to simulate the trial intend-to-treat analysis. So, in this analysis, we found the hazard ratio of 0.85, and the result was more significant because of course we have larger sample size. And this hazard ratio, 0.85, is very similar to the trial result, 0.86. And remember, our analysis were completed before the release of the trial finding. So when we conducted our analysis, we had no knowledge of the trial results, so we didn't manipulate our hazard ratios to make them look similar. But, interestingly, the results from our study look almost identical as the trial. This suggests that our analysis probably have good internal validity. For those who don't know, internal validity means a lack of bias, or confounding, and be able to estimate the treatment effects more correctly. So, clinical trials are often considered as a gold standard for estimating treatment effect. But in this study, our results in this large, real-world population, we replicate the trial analysis and we found almost identical results. This suggests that observational data, when conducted rigorously, can have very good internal validity and can very malleable in estimating treatment effects. But these data have complimentary strengths. In our intend-to-treat analysis, it's actually the best in terms of establishing causal relationship, because it maintains the randomization. But, the sample size was small, and we had some crossover between the two treatment arms, so we lack statistical power. And patients are not so representative of everyday practice. Because we mean, minorities and older adults are all under-represented in the trial. The per-protocol analysis is a little bit better in terms of statistical power, but it lost the randomization. In the sample size, it's still very small, and you can see the confidence interval was very large. On the other hand, the OptumLabs analysis cannot establish causal relationship, because our study is an observational study. We may have some residual confounding, and we don't have randomization. But, we had greater statistical power, and patients are very representative of everyday practice. So, in this case, clinical trials and observational data provide important complimentary evidence to help us understand this topic. And if you put together all these four analysis, you can see that ablation probably does have a benefit. And the lack of significant results in the trial intend-to-treat analysis is more likely due to a lack of statistical power, rather than lack of a true benefit. So, now, back to our three patients. In our first patient, William, he is eligible for the trial, and in such patients we found a hazard ratio of 0.7, which is very similar to the trial results. And that's about 30% relative risk reduction. But, interestingly, in practice the absolute risk was greater. So if you look at the left, in the trial at the end of five years, the absolute risk was about 10%, and the absolute risk reduction with ablation was about 2%. But if you look at our data in practice the absolute risk at the end of five years was about 15-20%, and we found an absolute risk reduction of 7%. So, we found a similar relative risk, but greater absolute risk reduction. This is not so surprising, because if you think about it, the control of hypertension, diabetes, and the use of statins or anticoagulants the use of all these drugs is lower in practice than in the trial. So, not so surprising that patients in practice often have worse outcomes and higher risk. So, although the relative risks are similar, the absolute benefit was greater in practice. For our second patient, James, James is not eligible for the trial because he's young and he has no risk factor for stroke. So the trial did not have any evidence for James. But in our data, we identify nearly 7,000 patients just like James. And we found a hazard ratio of 0.67, which is very similar to our primary analysis, about 30% relative risk reduction. But James' risk was so low, the absolute risk at the end of five years was only 2% versus 3%. So, in patients like James, we found a similar relative risk but smaller absolute risk reduction. For our last patient, Susan, Susan is not eligible for the trial either because she had renal failure, and she met one of the trial exclusion criteria. So she's not that good a candidate for ablation. And the trial did not provide any data for Susan. So, using our data, we found nearly 41,000 patients just like Susan, and in such patients, the absolute risk was so high at the end of five years, we can see the absolute risk was nearly 30%. But, the hazard ratio was only 0.85. That's about 15% risk reduction. And remember, in James and William, their relative risk reduction was nearly 30%. So, in such patients, we found a more modest effect in both the absolute risk and relative risk reduction. So, now back to our two treatment goals. For reducing cardiovascular events and mortality in patients like James, in patients like William, who was eligible for the trial, we found a greater benefit with ablation. But in patients like James and Susan, who are excluded from the trial, we found a smaller benefit. James is actually a good candidate for ablation because he's very young and healthy, he will do well on ablation. But, his cardiovascular risk is just so low. He probably just have little benefit from cardiovascular risk reduction. Susan has very high risk, but she's not that a good candidate for the procedure, so we found a smaller benefit in patients like Susan. But, remember, we have a second treatment goal. That is to improve symptom. And ablation is better than the drugs in terms of improving your symptom. So, for symptomatic patients like William and James, ablation may still be a good treatment approach. So, oh, so now... Who, for which patients would you recommend ablation? Of course, you can vote for more than one patient. Yes, very interesting, it's very different than the beginning, right? So, today we have been through this journey together to see how different types of data shape our treatment approach for individual patients. And we showed that our OptumLabs data can provide important complimentary evidence to help us understand how we should best care for our patients. However, these days, when people talk about clinical trials and observational data, there appear to be two types of people. One who really likes observational data, one who hates observational data. But our country is so divided right now, do we really need one more thing to divide us? And observational data and clinical trials are not in some sort of competition, and we do not need to vote for any of them. Instead, they are just different pieces of the puzzle. They together help us figure out how we should care for our patients. So what we need to do is not to argue which type of data is better, but rather how to better conduct these studies and how to integrate different types of studies. And today, we just show you one example of integrating those data, but there are many other ways to integrate them, right? For example, we are currently developing machine learning based prediction models in OptumLabs, and then we will validate those models in clinical trials. Then we will build these models into decision aid, to build these AI-powered decision aids. And this decision aid we will develop in practice and evaluate in trials. And, eventually, this decision aid will help us promote evidence-based shared decision making so we can work with our patients together to figure out how we can align both treatment benefits, align the treatment decisions with both treatment benefits and patient's goals and preferences. So by using all these innovative methods, and by integrating these amazing data sources, we believe we can revolutionize the way we conduct research and deliver healthcare. And I hope more of you will join us in this journey, together. Okay, thank you!