Wearable technologies may transform patient care, clinical research
Rapid advances in wearable technology have enabled patients to be monitored by their physicians as never before. These gains have ranged from enabling physicians to receive information about their patients’ heart rates and rhythms and activity levels from wrist devices (eg, Fitbits and others) and smartphone apps, allowing HF specialists to monitor their patients’ pulmonary artery pressures remotely, increasing the likelihood of detecting a deleterious change in their condition before hospitalization is required. The assumption is that such information will result in improved outcomes.
Cardiology Today convened a distinguished panel of experts to discuss the wearable-technology boom, how it is currently affecting clinical practice and how it might be harnessed to provide even more valuable information in the future, perhaps even affecting the design and execution of clinical trials.
Topics covered by the panel include how best to use data provided by current technologies, how such data can be individualized, the increasing push of “Big Data” into the medical community, the concept of passive data, the role of physicians and nurses in this new paradigm, and ongoing initiatives that could rapidly transform the way CV health is monitored and treated.
Targets for use
Carl J. Pepine, MD, MACC: My general impression is that there has been an explosion in use of modern electronic devices that are lightweight and wearable and can derive physiologic information that has the potential to be helpful to the patients and their physicians. Where do you believe this field is right now for cardiology and where do you think it can go?
Ilan Kedan, MD, MPH: For general cardiology, the applications aren’t as intuitive as one might expect based on advances in the area of congestive HF. With congestive HF, we have well-supported physiologic endpoints for treating, managing and diagnosing this disease. Except for the most obvious disease states in general cardiology, we don’t have as much experience with remotely acquired data for diagnosing and treating these problems. We are, however, always seeking to optimize care; for example, we can identify patients who have undertreated hypertension with remote BP monitoring and we can identify occult arrhythmias with remote monitoring. With activity data, we still don’t fully understand how to collect, analyze and apply that information to general clinical cardiology, but that is probably a future target.
Raj M. Khandwalla, MD, MA: What makes digital medicine so exciting is that these technologies will enable us to obtain completely novel data streams and new insights into human physiology. The next step is to develop effective strategies to apply these data streams to our patients, which will require well-designed, creative clinical trials. Take for instance, activity monitoring; for the first time we have objective data on how people move. This has significant implications for primary prevention and secondary prevention. We can now quantify how much exercise is optimal for patients. As computing power, bioengineering and connectivity progress, we’ll develop more sophisticated data streams from both implanted and wearable devices. In the near future, I suspect wearable devices will be able to provide continuous BP, bioimpedance, activity, pulse and arrhythmia data. As these data streams become more accessible, we will apply machine learning and higher-level analytic tools to predict CV events. In turn, clinical trials will be able to transform these data streams from theory to practice, leading to the next generation of medical advances.
Lynne Warner Stevenson, MD: In general, with the exciting new therapies with medications for HF, we are really looking to start individualizing therapy. We need to have real-time physiologic parameters to measure that are meaningful and actionable. Certainly, the first for us has been wireless pulmonary artery pressures. In addition to the fact that we can make a difference in how patients feel and what their outcomes are, it’s a whole new source of learning. We are learning things we never knew before, like what happens to people when they’re at home and not with us.
Javed Butler, MD, MPH, MBA: There’s absolutely no doubt that these wearable devices provide us with good data. With all of these nontraditional type of data that we get, my question is what will be our threshold for acceptance into clinical care? Are we going to put these things to test like we put pharmacologic therapies before we embrace them in clinical practice? There are multiple potential applications for these data, including management of patients, diagnosis, clinical-trial endpoints and, possibly, clinical trial management. The potential for their use is wide and exciting, but how to use them and the question of the threshold of acceptability are two things that I am completely unsure about.
Maria Rosa Costanzo, MD: I agree with Lynne. As we are collaborating actively in the area of pulmonary artery pressure monitoring, I would like to make a few observations regarding our experience. Surprisingly, we’ve found that patients adhere to taking their measurements much more than they adhere to their medications. We see patients who, for example, are not taking their diuretic as recommended, but they’re measuring their pulmonary artery pressures and transmitting them consistently.
Another area in which this is important is the interaction with the patient and using the data as teachable moments.
On the horizon
Pepine: Each of you has given a quick snapshot of where we currently are in this field. In your wildest imagination, what do you believe is on the horizon for wearables and CVD?
Costanzo: In a dream world, one device that I would like that doesn’t exist yet is one that measures intravascular volume continuously, without having to obtain multiple blood draws.
Kedan: There is great interest in what to do with these big datasets acquired with wearable devices. I anticipate a time in the future when there is a wearable device that passively acquires a meaningful dataset, that continuously streams, can be continuously analyzed, and then acted on. Machine learning of patient data may be accessed to personalize care as a predictor or an anticipatory model for events. Those events may not just be HF; they could be any number of different events, like blood glucose changes, asthma attacks or infectious events, in addition to primary cardiac events.
Khandwalla: Silicon Valley is always looking for new markets and the home is the next target. In the Valley, they call it the Internet of Things, and wearable medical devices are really a subset of this trend. The intersection of internet-enabled home and medical devices is quite exciting, and in the coming decades we may start to see people’s homes as an important part of the health care system.
So, in my wildest dreams, we have sensors on our body and in our homes that help us identify when we’re getting sicker or sadder even before we’re consciously aware of it. Then we can intervene upon ourselves and live the best lives we possibly can. It’s a brave new world fraught with both promise and peril, but one I believe will be a reality in the next quarter century.
Stevenson: My view of the future is a little shorter. The first thing I see is true triangulation of the data, where it goes in every direction but definitely goes back and forth between the patient and the monitor. So the patient becomes empowered to make changes that need to happen, and I become, after a while, primarily an observer and can intervene when necessary. I could even see that, in the future, we may be able to finally harness the power of biofeedback for a number of these cardiac parameters.
The next step into my not-too-distant future is for some things to have a closed loop — the same way we have diabetic pumps for insulin — for pulmonary pressures or for cardiac output in the future that we have medications that will titrate without necessarily external intervention.
Butler: Even now we have GPS-enabled apps that you can use in clinical trials when the patient goes to any ED, it automatically gives you alerts.
But I am, by nature, skeptical so I am not totally buying into the idea of Big Data yet. The technology is clearly ahead of where our science is, and I come back to my initial comment: What will be our threshold of acceptability? If something just makes sense to use, do we embrace it or do we put it to the test? We have particularly learned this with quality improvement efforts that make sense but when clinical trials are performed, these trials did not show any benefit over standard of care.
Khandwalla: But to think that the internet is not going to disrupt medicine is unrealistic. There is going to be a place and a time when information technology and Big Data have an impact on health care.
Stevenson: We’re going to have to have some novel trial designs. One of the things that was frustrating to us with pulmonary artery pressure data is that patients didn’t know we had it; what makes a huge difference is that patients know they are being monitored. So the idea of blinding in a trial removes a lot of the power of the intervention. We’re going to have to come up with something like center randomization, for example, or something where patients know that they’re being watched or cared for because that’s part of the power. We’re going to have to change our paradigm of what it takes. This isn’t like a placebo pill.
Kedan: Future devices might actually be seamlessly acquiring and uploading data so you could potentially test devices that were sham devices. For example, something like a patch, which doesn’t transmit data or transmits incomplete data, might be used for clinical research.
Another issue with the conventional strategy of clinical trials is that the devices, and the pace at which devices improve and accelerate in their capacity moves faster than our capacity to create, design, implement and execute a clinical trial. So, the device that you test and approve and demonstrate is effective may be obsolete by the time you’re done with the study.
Butler: Within the HF space, now we have clear randomized controlled trials of telemonitoring that demonstrate it does not improve outcomes. The investigators who did the trial will appropriately say, “Well, patients were not adherent and therefore the trial was not positive.” But if your intervention is such that patients cannot adhere to it long term, then it is unhelpful. Also, if you put these patches and devices on patients and there is clearly an initial upsurge of enthusiasm, the question arises, for chronic diseases, will it last for months and years down the road?
Khandwalla: To advance, we need better data streams. In HF remote monitoring trials, we are still using BP, pulse and weight, which we’ve been doing for the past century. We can’t expect to make progress with these archaic data streams. For instance, even though every first-year medical intern has been taught about the importance of titrating diuretics based on weights, we now know that weights aren’t all that predictive of HF admissions. When are we going to get past dogma and actually develop data streams that can improve our ability to treat CVD? CardioMEMS (St. Jude Medical) is an important first step. The CHAMPION trial was the first to demonstrate that you can use a data stream from an implanted device to improve outcomes. That’s huge and we need to build on that.
Need for more data
Pepine: It seems to me that the BP investigators have a clear need for better data. If we could sense increased BP and activate a device or drug patch to lower it, we could advance the field. We have pacemakers to increase heart rate, totally independent of the patient who doesn’t have any idea about what their heart rate may be. Are we moving in that direction?
Khandwalla: As Wi-Fi becomes more penetrated in every aspect of our lives — once our dishwasher and our thermostat become Wi-Fi–enabled, our medical devices will as well — and that provides a whole new array of opportunities for cardiologists and other medical practitioners. We need to embrace this trend and perform clinical trials with these devices. I think that’s an exciting future.
Stevenson: We have to be careful about the “Jeopardy game problem” where we are given a measurement by the industry because they can do it, and then asked to come up with how it might be useful. Instead, we have to ask the right question so that industry can give us the right answer that is actionable in the right timeframe for initial action and then reassessment. It is challenging enough to make the right intervention when we know the right target, without having to respond to surrogates that only correlate without actually measuring the important physiology that our therapies are designed to treat.
Khandwalla: What’s interesting is the philosophical disconnect between Silicon Valley, which uses a Big Data “shotgun” approach, and academic medicine, in which we define a clinical question and then develop a trial to answer it. On Wall Street and in Silicon Valley, they take large data pools, mine the data, look for patterns and then act. The verdict is still out on whether this approach can be applied to medicine. I suspect that it will be, but of course we need to be cautious. Advanced data analytics also brought us the 2007 financial crisis and failed to predict Brexit or the emergence of Donald Trump [in the U.S. presidential election].
Butler: So the issue is that the society will have to move to a different place than where it is right now if this were to work. One issue is dealing with HIPAA compliance. We don’t know how to deal with that and still be practical. A second issue is who will pay for the full-time equivalent staff that you have to hire to monitor data and intervene? Unless and until the data feed right back to the patient and then the patients are empowered to make changes in the medication, to constantly monitor all these data and to call the patients is not a trivial task. What happens on the evenings and weekends? What are the associated medico-legal risks? What are the reimbursement models? All of these points highlight fundamental changes that we will need to make. Data alone are not enough. Not that it is not possible, but it is not easy either.
Costanzo: One thing that you mentioned is people, and the manpower needed to review and react to the data. I had to have this discussion recently with nurses who work on my team. I walked them through the process that is happening now vs. the process with the patient that has a pulmonary artery pressure monitor because they immediately interpreted the device as additional work for them.
Stevenson: I want to make a couple comments again about Big Data, which is complicated by “Big Noise.” One of the concepts that requires scrutiny is the idea that if we knew everything today then we could predict everything tomorrow. Often there is a completely unpredictable factor that tips everything over. The more direct the physiology we monitor and the closer to real-time that we can see it, the more we can steer it back on course. But, in fact, as it has been very clearly shown by engineers, engineering modeling does not translate well to biological systems. You may be able to predict when the screw is likely to break, but human biology interacting with mechanical factors includes many more random elements.
Costanzo: We learned that in hard ways with continuous flow left ventricular assist devices.
Kedan: It’s important to understand what the limitations might be and appreciate an achievable or realistic goal for what you might do with these wearable data. If it tracks well with a patient and we can learn that from 6 months of care with a patient, we might be able to decrease the hospitalization rate by 10% or 15% or 20%. Every day spent in a hospital is expensive. So for every day, you have increased vigilance in managing these more frail, elderly, sicker, multi-disease patients. If we can move that needle to decrease hospitalization, or decrease acuity of care, it might be worth it to invest more in these patients with wearable remote monitors.
Stevenson: I agree completely, but what I’m making sure we’re careful with is this concept with Big Data. When we put in hundreds of numbers that we don’t understand individually, and then trust the machine to learn what it means, the question is what do we do with the output? “OK, now you can predict this, that or the other.” If the prediction is not transparent, then what is the appropriate action? We are fooling ourselves if we think that we just plug in to Big Data and wait for the answer.
Butler: One thing that we’re not talking about for more monitoring is the remote visit for people living in rural areas. There is plenty of good that can come out of it, but coming to Lynne’s point that this Big Data will not somehow aggregate all these pieces of information into automatic actionable items. If the promise for Big Data is that it will make doctors and nurses better providers, absolutely yes. But if the promise is that this will automatically generate care pathways and then the patients will take care of themselves and you won’t need doctors and nurses, I will be the first one to say that that day will never come.
Khandwalla: I agree that in the near future computers will not tell us how to take care of patients, but they will become increasingly helpful in diagnoses and treatment. Right now, Big Data will help us identify potentially new avenues for clinical studies and investigation. As clinicians, we need to look at these new data analytic tools and large datasets to determine what questions are worth asking and then design good trials to answer them. That’s something we do in cardiology all the time irrespective of the size of the dataset.
Costanzo: We need proactive physicians and nurses because we cannot forget, for example, that a pulmonary artery pressure monitor is a diagnostic device. If the physician does not react to the values that are obtained by the device, the device is not going to help us take care of the patient. Our most arduous job is to create that mental shift into the physician providers — that they have data but they need to do something about the data in order for the data to improve the patient’s outcomes.
Current use of wearables
Pepine: Let’s focus now on what you use in your patients currently that can help in their care.
Costanzo: I’ve been involved with implantable hemodynamic monitors since the early ’90s. CHAMPION, the trial with a pulmonary artery pressure monitor, was the first one that led to positive outcomes. The specific reason, in my view, is that compared to previous trials, the mental shift finally occurred that we had to react to this number, to this target, not to symptoms.
Pepine: If you had your way, do you believe that continuous pulmonary artery pressure monitoring would be something that you would wish for in your patients?
Butler: When we talk about this type of device there are four kinds we are thinking of. One is the implantable device. The second is the patch that you wear on your skin. The third is other wearables like an accelerometer. The fourth are apps that can go on your cellphone and capture data. Today, the only thing we have is the CardioMEMS device. There were other types of monitoring devices, but for various reasons trials were not as positive with them as with CardioMEMS. That is the only clinical application in patients with HF that we have today shown to improve outcomes.
However, there are other applications that are helpful as well. For instance, you can have a reminder system for compliance with medications and all those kinds of things that you can work on. Clearly, there is a need and an interest in novel clinical trial endpoints using these mobile app devices. But in terms of management, we are not far beyond the implantable pulmonary pressure monitor at this point.
Stevenson: There have been a number of devices related to impedance, both major externally and internally, and I have been thwarted in trying to put these to good use. First, if a threshold is going to be actionable, it has to be extremely reliable. And it needs to be actionable for what you can do. If you pick up the fact that they’ve got a cold or bronchitis and you treat them with diuretics, that’s a problem. The second issue is that it has to have the right kinetics of response, so if I have not intervened enough I have enough time to revise the plan.
Costanzo: The CHAMPION trial with the pulmonary artery pressure monitor began in 2006. Having enrolled some of the early patients, and having performed right heart catheterizations on them 10 years later, I found that the CardioMEMS monitor measured pressures with the same accuracy as at the time of implant.
Stevenson: What I’m using now is the pulmonary artery pressure monitor. Other things that would be helpful would be very convenient activity monitors so I could see that patients are actually doing what they tell me they’re doing. What’s available now that’s useful also are the rhythm metrics. For example, percent of atrial fibrillation, percent of CRT pacing and so on are very important. We know that if we get under 95% of CRT pacing, we need to intervene to improve that. Very selective things from the rhythm devices have been helpful, but many of the parameters provided have been like the Jeopardy game again.
Khandwalla: In our practice we are actively trying to get as many patients onto Apple HealthKit as possible. Cedars-Sinai has a relationship with Apple that allows us to sync HealthKit data onto the electronic medical records. Activity data show up under one of the tabs. We don’t yet know what we are going to do with the data, but we hope in the future we can ask important clinical questions like: If patients average, say, 8,000 steps a day for 5 or 10 years, what’s their risk for developing obesity?
We’ve also tested a wearable necklace, with data that I presented at the American College of Cardiology Scientific Session earlier this year. The toSense CoVa Monitoring System is a novel necklace that measures impedance, and also has very high correlation with cardiac output and stroke volume in 24 patients compared to cardiac MRI. We did a 20-person pilot study that showed it has the potential to predict HF events, although it still needs to be validated in a larger population.
The third thing that we’re very interested in is using wearable technologies in clinical trials. We’re working with your research team, Carl, and investigating the efficacy of novel medications in microvascular coronary disease. We’re working with C. Noel Bairey Merz, MD, using Fitbits to test whether ranolazine (Ranexa, Gilead Sciences) actually increases patient activity in participants with microvascular coronary disease.
Kedan: One thing we need to do better is understand the effects of our therapies. This is not just our medication therapies, but surgical interventions, or any other physiologic change that we may enact on a patient; what is that doing? We have survey scores and we have subjective assessments of patient change. We need better data. So an activity data, some sort of accelerometer data, heart rate data, BP data, oximetry data, that can all be used to better help us understand what therapies are actually working, and what therapies will work. In the setting of postcardiac surgery, for example, how many steps is enough or too many to take during recovery? We need to know what the right amount of activity is so that our patients can improve optimally and that we can tailor or personalize the therapy for each of our patients.
Butler: There is potentially a great future with this. We will get a lot of data that can potentially make us smarter and potentially make us better providers. But we have to be circumspect in how we will test the utility of those data and what societal changes will we need to make within health care, within our patient communities and within the regulatory and legal environment that we live in for that to come to fruition.
Khandwalla: There is a lot of optimism, and that optimism needs to be tempered by the scientific method. Ultimately, it’s going to be a really interesting test to see if this Silicon Valley approach actually works. I think it will ultimately require clinicians to work very closely with technology companies to develop clinical trials. Eventually, I’m optimistic that Big Data will help us find useful signals to improve cardiac care.
Costanzo: I’m not objecting to this Silicon Valley approach, but instead of collecting the data first and involving the clinician later, you need to involve the clinician up front so that you collect relevant data.
Khandwalla: Two weeks ago, one of my favorite patients was found dead in a chair at the age of 67, 2 days after I saw him in clinic. He had coronary disease and AF. He was on optimal medical therapy, his BP was controlled, and his EKG and ventricular function on echo were unchanged from previous assessment. As I was talking to his son, I kept thinking to myself: Was there something I could’ve done to prevent his death?
That’s why this so important. I hope one day a device like the AliveCor Kardia Mobile ECG can alert me to my patient’s impending demise. That’s what I’m focused on and ultimately it’s going to be up to us as cardiologists to use this new technology to help our patients live better and longer lives. I’m confident we can do it.
Stevenson: There’s no question that we’re moving into the future. The future is going to involve all sorts of electronic advances. We need to distinguish carefully between what is the best way to collect information that we know right now is valuable and actionable, and at the same time be alert to discovery events, with Big Data to try to find the physiology that we’ll act on in the future.
Pepine: I thank all of you for these valuable comments. From my position, the take home for our readers would be keep your eye on the next generation of wearable sensors. They will be capable of biological and chemical sensing to improve care and reduce overall costs of health care delivery. The technical challenges are attracting solutions and future trends will be eye-opening. They will measure multiple human performance metrics continuously. Development of a novel class of minimally invasive electrochemical biosensors will facilitate quantification of relevant metabolomic, electrolytic, hormonal, and neurochemical information in a continuous, real-time manner. Novel manufacturing processes are yielding unobtrusive, low-profile, skin-applied devices able to deliver timely, clinically-accurate and actionable information using existing wirelessly enabled wearable and mobile platforms. Continuously monitoring both physiology and biochemistry of our patients, in their native environment, has the potential to dramatically impact health outcomes in a favorable direction. – compiled by Katie Kalvaitis
- Abraham WT, et al. Lancet. 2011;doi:10.1016/S0140-6736(11)60101-3.
- ClinicalTrials.gov. CardioMEMS Heart SensorAllows Monitoring of Pressure to ImproveOutcomes in NYHA Class III Heart Failure Patients(CHAMPION). www.clinicaltrials.gov/ct2/show/NCT00531661. Accessed on Oct. 3, 2016.
- Khandwalla RM, et al. Predicting heart failure events with home monitoring: Use of a novel, wearable necklace to measure stroke volume, cardiac output and thoracic impedance. Presented at: American College of Cardiology Scientific Session; April 2-4, 2016; Chicago.
- For more information:
- Javed Butler, MD, MPH, MBA, can be reached at 26 Research Way, East Setauket, NY 11733; email: email@example.com.
- Maria Rosa Costanzo, MD, can be reached at 801 S. Washington St., 4th Floor, Naperville, IL 60540; email: firstname.lastname@example.org.
- Ilan Kedan, MD, MPH, can be reached at 250 N. Robertson Blvd., Suite 403, Beverly Hills, CA 90211; email: email@example.com.
- Raj M. Khandwalla, MD, MA, can be reached at 250 N. Robertson Blvd., Suite 403, Beverly Hills, CA 90211; email: firstname.lastname@example.org.
- Carl J. Pepine, MD, MACC, can be reached at Cardiology Today, 6900 Grove Road, Thorofare, NJ 08086; email: email@example.com.
- Lynne Warner Stevenson, MD, can be reached at Brigham and Women’s Hospital, Division of Cardiovascular Medicine, 75 Francis St., Boston, MA 02115; email: firstname.lastname@example.org.
Disclosures: Butler and Pepine report no relevant financial disclosures. Costanzo reports receiving a research grant to her institution for the CardioMEMS PMA trial and honoraria for participation as a steering committee member of the CardioMEMS PMA trial. Kedan reports receiving a research grant from General Electric. Khandwalla reports financial ties with General Electric, Novartis and Sentrian. Stevenson reports receiving research support and serving on an advisory board(s) for St. Jude Medical, serving on an advisory board(s) for Medtronic and receiving research support from Novartis.