The Patient Equation

Home > Other > The Patient Equation > Page 7
The Patient Equation Page 7

by Glen de Vries


  That said, I don't want to undersell the exciting devices and initiatives out there—from “smart” everything to ingestibles and implantables purporting to change our lives and our futures. It is not hard to get excited about the digital world we're living in—and even more excited about the promise of the digital world to come. As sure as I am that these technologies alone won't power the future of health, they will be a part—an essential part—of it.

  Are Apps the New Snake Oil?

  An app called Cardiogram, available on the Apple Watch and other devices, sends heart rate and step count data to a machine learning algorithm they call DeepHeart, which aims to predict sleep apnea, high blood pressure, early signs of diabetes, and more.3 Its atrial fibrillation screening has been shown to be 97% accurate, and it has been able to stratify patients into risk groups for cardiovascular issues.4

  Babylon, based in the UK, has claimed to be able to improve on WebMD, and in fact uses artificial intelligence within its app to diagnose patients even better than a doctor can. “I don't think it is going to be as good as a doctor,” Ali Parsa, Babylon's founder, told the Financial Times.5 “I think it is going to be 10 times more precise than a doctor. No human brain is ever going to be capable of doing anything of the sort.” (Unsurprisingly, that claim has been disputed in a study published in The Lancet: “Babylon's study does not offer convincing evidence that its Babylon Diagnostic and Triage System can perform better than doctors in any realistic situation, and there is a possibility that it might perform significantly worse.”6)

  Another app, Healthymize, listens to a user's voice and breathing patterns while talking on a smartphone in order to detect signs of chronic obstructive pulmonary disease and alert users and caregivers.7

  One more app, Migraine Alert, is designed to predict migraines before they happen, collecting a set of potential triggers (weather, activity, sleep, stress, and more) and using machine learning to create personalized predictions—with 85% accuracy after logging 15 episodes for a particular migraine sufferer.8

  These apps, and countless others like them, may or may not prove to have clinical value over time. It's one thing to be able to predict a migraine three days in advance with 85% accuracy9—it's quite another to be able to help patients actually do something to prevent the migraine, guide them toward treatment, or change something about their behavior or environment in order to make fewer migraines happen for them over time. Machines generating information can sometimes seem like enough to impress—but unless that information is useful or actionable, it's hard, once you get past the initial “wow” factor, to see the point.

  Skepticism also requires asking the question of how often the subjects in the study had a migraine. If they get them six days a week on average, and we guess they get one every day, we will be right 85% of the time. The point isn't to poke snarky holes in research like this. The point instead is that we need to be scientifically responsible and make sure we are creating useful predictions, paired with—or at least leading to—useful preventions and interventions. We need systems that are able to connect the dots between the data and the larger understanding of disease progression, creating beneficial outcomes for patients. We need more than just the raw data. And we certainly need more than product marketing.

  Wearables for Panicked Dogs

  It's not just apps out there. A recent annual Consumer Electronics Show (CES) featured over 250 exhibitors displaying new products across the wearables category—from tech‐enabled rings and sneakers to pajamas and pet collars.10 One company has developed what they claim is an emotion‐sensing wristband that can tell you whether you're happy, sad, or stressed.11 Another claims to have created a sensor that can detect panic attacks—which is coupled with an app that can help guide you through them.12 Combine the emotion sensor with the pet collar, and an entirely new market might emerge, tracking the inner lives of our pets.

  Not long ago Apple made news by filing a patent application for a new wearable device—a piece of jewelry capable of performing an electrocardiogram beyond what an Apple Watch can do, generating accurate readings regardless of where it is worn on the body.13 At Northwestern University, researchers have developed a sweat‐analyzing skin patch that can monitor glucose levels and other indicators of health.14 And Google's life sciences company, Verily, has expressed a desire to make bodies produce streams of telemetry like race cars—by instrumenting people with glucose‐sensing contact lenses, handing them tremor‐canceling silverware, and more.15

  It is not impossible to imagine a future world where the devices truly make a meaningful difference, where people have 3D medication printers at home producing custom supplements for them every morning, built off data generated by the waste collected in their smart toilets, optimizing what they need for the day. It is not impossible to imagine implanted or ingested devices measuring temperature, heart rate, and blood chemistry, operating quietly in the background of our lives, not requiring any particular attention from us, sending data into the cloud that will send back prompts and recommendations.

  But we're not there yet as far as clinically‐validated usefulness, not even close. And the line between fashion and medicine, between glitz and medical value, is fuzzy, to say the least—and getting fuzzier all the time. It is easy to dismiss all of this as a fad, or at least as a set of products that exists in an entirely different business space than the traditional field of life sciences. We all know about the failures of companies like Theranos, which raised more than $700 million promising breakthrough technology in blood testing that turned out to be built on a platform made of quicksand.16 IBM's Watson supercomputer was supposed to be able to use data to cure cancer, but ended up recommending unsafe treatment plans.17 And that glucose‐sensing contact lens that Google was developing? After much hype, the project was summarily cancelled.18

  In reality, even the best wearables have issues. One is that people just don't use them. According to one study, a third of people who buy “smart clothing” stop wearing it.19 Significant numbers of people wearing fitness bands, smart watches, and smart glasses simply find them uncomfortable, unstylish, or inconvenient to charge or sync with their smartphones. (That's a problem we can perhaps overcome as more actionable health information emerges from these devices. It's one thing to stop wearing a tracker because you stop caring about your step count. It's a different issue if instead of step count, it's telling you whether or not your chemotherapy is working.)

  Battery technology is probably the biggest barrier to compliance with wearing devices consistently and getting good data from them. But this issue is being worked on by industries beyond just wearable devices. Whether we're talking about our cars, or homes, or our wearables, batteries are getting better at holding larger charges and for longer periods of time. The electronics they power—from motors to microchips—are getting more efficient at using that power. Wireless charging continues to grow, and there are promising technologies for making it work at greater distances. So at least with regard to the wearables that connect to our patient equations, it's probably just a matter of time until the issues of charging are solved. The material science and electrical engineering is being taken care of. But we also need to look at other issues.

  A report in PLOS Medicine, titled “The Rise of Consumer Health Wearables: Promises and Barriers,” explains that despite the hype, there is no evidence that wearables impact behavior, let alone improve health.20 “Many wearables suffer from being a ‘solution in search of a problem,’” write the researchers. The people drawn to using them are already healthy, they don't stick with the wearables long enough, and the data produced hasn't been proven to mean much in the first place.

  Finally, another issue is the problem of data misinterpretation. One study in Sweden has shown a correlation between low resting heart rate and a propensity for violence.21 But causation isn't the same as correlation, and if that information were to be used to profile individuals or potentially even convict someone of a crime, the
n it would seem that we would certainly be taking a huge and unjustified leap from the data. An article in the online magazine Slate wonders about the implications of data being used in the workplace: just taking sleep data, for instance, and the idea that a well‐rested employee will produce better work, “imagine if your manager began making decisions about which projects to assign you based on whether you were well‐rested.”22

  I spend lots of my time on the road, talking to people on all sides of the life sciences industry, and they all tell me versions of the same story. People have been talking about precision medicine for years now, and where has it led us? Down a lot of blind alleys, chasing magical cures that turn out, upon real analysis, to be largely mirages. Clinical trials have been running the way they do for generations—and there's a reason for it.

  My response? Yes, not every prediction of technology‐powered industry transformation has come true, sure. But that doesn't mean we haven't made incredible progress, and that we're not on the cusp of true transformation. Even with the unanswered questions and doubts, we shouldn't make the mistake of throwing the baby out with the bathwater. It's fair to be a skeptic—indeed, it's smart to be a skeptic, until science proves otherwise. But if you're not seeing the potential behind the headlines, you're at risk of missing out on the undeniable future. I am as much of a skeptic as anyone—and this book will talk later about how wearables and smartphone apps ought to be subjected to the same regulatory approval as prescription drugs before they're deployed to the public—but I also wear my chest patch and wristband. It's not that they're changing my life—yet. But there's a real error to be made in assuming that the underlying potential isn't there, and that achieving that potential isn't closer than it might first seem.

  It's the Equations, Not the Devices

  What the glitz and the headlines allow us to forget is that it's not about the devices themselves. It's not about the apps, and it's usually not about the front‐end technology at all. It's about the systems behind the scenes—and that's where the attention needs to be. Even the fanciest, most futuristic‐sounding new devices are largely built on sensor technology that has been around for years—decades, even. The accelerometers in our phones and smart watches use the same piezoelectric technology—the ability of certain materials to generate an electric charge in response to applied mechanical stress23—that helped put people on the moon. Step counting, sleep tracking, measuring your pulse by shining a light onto your skin—these may be new applications, but the underlying technologies date back to the Eisenhower administration.

  There are absolutely exciting new sensor technologies—measuring tiny metabolic changes in cancerous tumors or continuously monitoring blood chemistry via a patch on the skin, just to name two—but the most exciting sensors for extending and improving life right now are available on Amazon and have been (albeit for a long time in much larger physical packages) on the market for years. The game‐changer isn't the wearables themselves, but the algorithms behind the scenes, armed with sufficient data and computing power to do the real work needed to draw powerful connections.

  As we proceed into the rest of the book, we have to think not just about the current state of the science, but about the promise of the future. It is entirely realistic to expect that these algorithms may one day soon be able to take the data from the devices I'm wearing—combined with genetic data or other inputs in those multivariate equations we've already talked about—and tell me if I'm on a course toward depression, Alzheimer's disease, cancer, or more. They may catch virtually imperceptible movements in my hands, setting off a virtual alarm that harkens the possible onset of Parkinson's disease. Other devices might help detect pollutants in one's home and guide their doctor to order blood work, or use sleep patterns to tune the dosage of heart medications. These devices might not just measure a user's resting heart rate, but also notify them when it's optimal to exercise and how much exercise to do—based on metabolism, medical history, and environmental conditions—and whether or not a cheeseburger was eaten earlier that day.

  With millions of people around the world carrying smartphones in their pockets or wearing them on their wrists, collecting data at mass scale, it's shortsighted to deny the potential. The smartphone is a game‐changer for medical research in much the same way that reliable Internet was a game‐changer for the entire technology industry. When we started Medidata more than 20 years ago, our premise—that clinical trial data could and should live in the cloud, accessible to everyone on all participants in the process—was questioned by many. What if a doctor didn't have a computer? What if the Internet line wasn't working? Our key insight was realizing that an inflection point was coming so very quickly. Doctors were going to have computers. And if the Internet wasn't working, they were going to get it fixed as soon as they could, because they weren't just going to be using it to upload data from their clinical trial—they were going to be using it for medical records and billing and to tell their family when they'd be home for dinner. In the span of less than a decade, we went from Internet connectivity being something we had to seek out to the Internet being ubiquitous in our lives.

  It's been the same thing with smartphones. The proliferation of connectivity proximal to everyone's body is what has created an inflection point in the world of sensor data. Accelerometers have been around for decades. They've been in medical devices, pedometers, and toys. But all of a sudden, they became part of the infrastructure in our pockets, and we can measure people's data continuously.

  Although I certainly wasn't privy to any early design conversations about the iPhone, all the available information points to the incorporation of accelerometers into the device in an attempt to conserve battery life. Once again, the battery—and its limitations—rears its head, this time not as a limiter to how compliant a patient might be with a wearable, but as a catalyst for innovation. Just as Steve Jobs demonstrated, the earliest iPhones would turn their screens off whenever they weren't needed, for instance, when the phone was against the user's face for a call, or placed on a tabletop.

  Was it the case that the incorporation of the battery‐conserving accelerometer was part of a grand scheme to revolutionize the world of medical devices? That's possible, of course, but the simplest—and more plausible—explanation is that it was only after the technology was already in the phone that people realized the medical implications, for example, of counting steps with the accelerometer. Honestly, we haven't even figured out all the potential measurements we can gather, just based on the sensors we have. Withings—formerly Nokia Health—uses an electrical sensor measuring resistance as an indicator of cardiac health. This application wasn't even on anyone's radar screen a couple of decades ago.

  Of course, measuring how many steps we take each day is not an interesting fact on its own—until and unless we are able to discover that the number of steps someone takes in a day is somehow correlated with a medical outcome we're interested in, like the tumor burden in a growing cancer. At that point, depending on how closely those numbers track, it may become hugely valuable to throw a Fitbit on a cancer patient, especially if it gives us the possibility of seeing that increasing tumor burden more quickly than we could otherwise measure it in any realistic way. This is where these devices enter the patient equation.

  Similarly, a smart toilet analyzing what comes out of our body may sound like satire, until and unless we're able to discover that there's some biomarker we're able to measure that tells us whether or not someone is going to be a good responder to some new probiotic, or if there is predictive power in the relative abundance of species in our microbiome. Beyond the glitz of these wearables and smart devices is the real promise of helping us to better understand disease—but the way we're going to find that out is through disciplined research, including clinical trials. There is no shortcut to this process. We need to do the hard (and often expensive) work of testing those connections and figuring out what really matters when we're looking for the inputs that are useful
for measuring and predicting patient outcomes.

  This is where the life sciences industry can come in. It may be that step counts tell us a lot when we're looking at tumor burden but nothing at all when we're looking at who's on a road toward Alzheimer's disease. Or maybe it's the reverse. There's an industry that is spending the money and putting the effort into measuring and understanding the progression of cancers and neurodegenerative diseases. Drug and medical device companies are already running research projects with a rigorous scientific and regulatory framework, and given the minimal incremental costs of these digital measurements, there is a tremendous opportunity to incorporate them into existing research projects and to begin to answer these questions. These answers are the holy grail—not the devices, not the apps, not the newest press release about smart glasses.

  What's happening in the world of sensors is exciting, but what is most exciting is the science of using that sensor data to define, measure, and create mathematical models of disease that can lead to better outcomes—and huge benefits for everyone in health care and life sciences. Doctors are going to have more successful outcomes with patients, pharma is going to have more targeted and more successful drugs, and payers are going to have more success in managing huge numbers of people. Patient equations are going to help us uncover breakthrough new drug approaches, epidemiological discoveries that are going to change health across populations, and new ways to engineer clinical trials that will bring us into the twenty‐first century.

 

‹ Prev