The Silicon Valley Health Trend Making Doctors Nervous

0:00Today, opportunity could begin here and take you anywhere. At HSBC, we're here as your banking partner to help shape what's next. And we're there to help you navigate new markets from Seattle to Singapore, London to New York. HSBC brings strategic insights and a global perspective to give you an edge wherever you go. HSBC, we're here. So, Rachel, a little while ago, you got into a little spat online. Tell me what that was about. Yeah, so a few weeks ago, a startup in San Francisco announced that it was going to launch a new product, which is a whole body imaging technique. And it led to this really interesting, kind of nasty, it's fine, discourse between medical doctors like me and folks in tech.

0:57And that sort of conversation came down to doctors being very wary that this kind of data can sometimes cause more harm than good. And folks on the tech side saying basically, how can more data ever be bad? Isn't more data always helpful in informing decisions, research, et cetera, et cetera? And it led to this conversation between you and I actually about is more data about your body always good?

1:28That's the question that we're going to be talking about today in a bunch of different ways. It's not entirely new. People have been trying to track data about their body and their health forever, in more systematic ways, maybe over the last 10 years. But the prospect of AI really kind of changes the landscape here and makes us think again about what the future holds. Is it science fiction to imagine that there will be a day when an AI could predict 20 years in advance when a person is staring down the barrel of a neurodegenerative disease and act at a time when maybe we could actually reverse it.

2:01We're now looking at a future where many people are telling us that machine learning can process huge amounts of data much more quickly, much more intelligently than anyone has before and essentially learn things about how we are living, what health is, what illness is, and how we might be able to do better to manage our health going forward.

2:24I'm David Wallace-Wells. I'm a writer for New York Times Opinion and a columnist for The Times Magazine. I'm Rachel Bedard. I'm an internal medicine doctor in Brooklyn and a contributing writer at New York Times Opinion. So before we talk about the very now and the kind of distant future, let's talk a little bit just about the recent past. Over a couple of decades, doctors have become a little bit more skeptical than I think the average layperson about the possibility that knowing more is always good. Tell me about where that came from, what that suspicion arises around, and what normies like me don't understand.

3:03Yeah, so I think there are a few factors here. One is, you know, over the last several decades, our ability to collect data about the body has taken off radically, right? So imaging techniques, blood tests, tracking devices, all of these things can provide so, so much information that may or may not correlate in any meaningful way to what people feel in their bodies, how they're functioning, their clinical outcomes. When you're talking about studying people who feel healthy to see if actually they might be sick, you're talking about screening a healthy population for hidden pathologies. And there have been lots and lots of studies over time to try to do this.

3:48And what we have found is the results are really mixed. So the most sort of famous cautionary tale is South Korea in the early part of this entry instituted a policy where they were screening universally for thyroid cancer with thyroid ultrasounds. Thyroid ultrasounds, non-invasive, they don't cause any harm to do. Fine. What they found, though, when they followed that experiment over time was that the incidence of finding thyroid cancer went up 15 times and it made no difference to mortality from thyroid cancers. So which means like you were basically finding 15 times more cancers that weren't actually clinically significant, that weren't going to hurt people.

4:32And that's. Well, let me just pause you there. Like, how how is that? How can that be? How can that be? I mean, I understand, like, there are some things that are below a clinical threshold, which maybe we don't want to worry about. But how can it be that when we see are seeing so many more cases of something, it doesn't have any population level benefit? It could be in two ways. One thing is that there are just absolutely like indolent cancers that can sort of exist in small, very, very, very slow growing ways that are just not going to ever become clinically significant in a person's lifetime.

5:02Like prostate cancer, there's sort of this old adage that like more men die with undiagnosed prostate cancer than get diagnosed with it in their life. Yeah, I've heard people say we shouldn't even talk about it as a cancer. We should treat it as something else because the word cancer scares people into treatment. So that's one thing. The other thing is whether or not screening when somebody is asymptomatic is actually useful, right? So it may be that if you wait until people's thyroid cancer becomes clinically significant, until it's found because they have symptoms or on an exam of their neck, if you wait until then and intervene at that point, it's fine.

5:41The vast majority of thyroid cancers are caught pretty early. And when they're caught, they're very treatable and people do really well. And so there may be no additional benefit to catching them way earlier. It doesn't, that example, that cautionary tale does not mean that sort of we've closed the case on like, how should we look for thyroid cancer forever, right? But it means given the screening technique that we know how to use now, applying it at the population level to asymptomatic people seems to have no clinical benefit and instead causes a fair amount of harm because that 15-fold increase in cases means follow-up surgery, biopsies, surgeries, and all of those things.

6:24One big question that I have about this, not just about thyroid cancer, but the question of, you know, what we can learn about the body in general, is if we look at the state of play now and we say, given the treatment techniques we have, given the screening techniques we have now, expanding our screening to the whole population isn't going to have a benefit. Is that because we already know everything there is to know that is useful about such a disease or other diseases? Or is it something about the limits of our screening? In a future where we could zoom down, you know, have much more information about particular cancers, presumably more data would be good, right?

7:02It really matters if you know what you're going to do with the data. Okay, so let me give you an example that's a more live question, which is screening for Alzheimer's disease. Alzheimer's disease is incredibly prevalent, clinically devastating, on the rise, right? And until relatively recently, we had very few interventions to offer people if you knew that they were at increased risk for Alzheimer's. we didn't have anything that made the disease slow down or reverse its course. In the last 10 years, we have both found new screening techniques, blood tests that can find sort of evidence of early plaques in the brain, basically, that are developing well before you have any clinical symptoms.

7:57And the other thing that's happened in the last 10 years is for the first time there have been new approved treatments for people who are in very early stages of Alzheimer's. That's a total game changer because in that case, if you'd had that blood test 20 years ago and you didn't really have anything to offer people, the rationale for screening would be really low, right? Because you would say, you're just going to tell people this. They don't know that for sure it means they're going to get Alzheimer's, but it maybe freaks them out for the rest of their lives. Alzheimer's is in my family. I would never have gotten the screening test 20 years ago.

8:32It's really different if you have an intervention to offer people that may be meaningfully disease modifying. And so the question of whether screening is useful and the data is useful also goes in tandem with like, what are you going to do with the result when you get it? So what's the big problem with this particular full body scan that we're talking about today? Like, why is this an example of something that is going to give us information that is not useful, maybe counterproductive, as opposed to helpful to the people who are getting it? Yeah. So embedded in that question is like sort of the whole thing.

9:08Yeah, okay. So let's unpack. Let's unpack it. So the first thing is like when you talk about it being helpful to the people who get it, it really depends what the person's getting it for, right? If you're like, I don't know, Joe Rogan, and you're a fitness-obsessed gym bro who is working at several hours a day and really obsessively tracking, you know, your diet and all of these metrics about yourself, and you want to collect these images because you want to be able to see the relative proportions of muscle-to-body fat in your body, whole-body ultrasound is probably okay for that. and if that's something that Joe Rogan finds like meaningful on a personal level to himself like he's like this makes me feel better about the way that I'm taking care of myself go with God, Joe Rogan, enjoy, you know and that's sort of what the company is saying right now the company is saying this is not for medical use this is like a general wellness thing that people can use in order to track their body composition that's though a really different prospect than the way in which this conversation about it online was sort of extrapolating the potential benefits of such a technique, which were like, you're going to be able to get a monthly scan that will track the appearance of abnormalities in the body that may or may not be clinically significant and make sense of them.

10:30And that becomes a problem for a few reasons. The first is that we know when we scan people that we're finding stuff in their bodies all the time that we don't know what it means. We call them incidentalomas because they're incidental findings that are like of totally indeterminate significance. We're always finding like schmutz on people's adrenal glands. And there's all this, you know, there are guidelines about like, how big does the schmutz have to be for you to decide that you're going to scan again and which interval, et cetera, et cetera. But that stuff's meaningfully costly and potentially harmful because you have to pay for scans and with time and money, you get biopsies, you know, all of these things that are potentially harmful without any benefit.

11:13So that's one reason that it makes doctors really nervous. The other reason I think that it makes me really nervous is because this is sort of part of this like larger trend of direct to consumer access to medical testing or what is sort of like medical testing adjacent, right? There are also companies that like, you know, where you can be ordering your own lab panels and then getting back all of this blood work. And whether or not that's meaningful data about your health is sort of hard to say. And once you get it back, when there are abnormal values, like your next step is you're taking it to your doctor and saying, this seems to say, you know, these values are off.

12:00Like, what am I going to do about it? Well, some people are taking it to a doctor, but also a lot of people are just monitoring it themselves, right? Well, monitoring or taking action on it themselves. Like, what's that? Right. Like, what are they doing? And there's a conceptual shift that's happened where, you know, previously they had sort of assumed that they were in relatively good health. They start to see some indicators that may or may not mean anything, but they've already stopped thinking of themselves as being in good health and started thinking of themselves, if not unhealthy, then on some spectrum of wellness and performance.

12:30Totally. They maybe should be doing better. They should be addressing this or that. And whether or not those improvements will actually help their well-being in the long run, they're already mindful of what they could or should be doing. So they've already, like, redefined their measure of wellness from, like, how am I feeling to what does my watch say how I'm feeling? Totally. Totally. So in preparation for our conversation today, I have been wearing for the first time in my life, a fitness tracker, like a sort of watch type device for the past week. You're like really a late adopter. This, I'm a really late adopter and I'm also.

13:07Although I'm a Luddite, I don't have anything. You don't have anything. It's just vibes in David Wallace as well as his body. I'm doing great. Not me. I've been tracking my data for a week and I cannot make any sense of it. Last night, It told me I had a bad sleep, but I felt I had a great sleep. And I really did, like, look at the data this morning, and I was like, well, what does it know that I don't know about what was happening? You did have that feeling. You didn't have the feeling of, like, I know better than this watch. Well, I was like, I feel pretty good. Two nights ago, I slept terribly.

13:35And the watch, the store thought it went fine. And then this morning, the device thinks that I slept badly, and I woke up feeling much better. and mostly I'm going to defer to my own experience, but I did like kind of look at the graph to be like, what does it know that I don't know? I mean, there's data that's come out that says like, you know, there's a placebo effect and a nocebo effect to all of it, right? Which is like, if the device suggests to you that you had a bad sleep, people experience more tiredness that day, whether or not it's true. Yeah. I mean, it reminds me a little bit of some of the conversation around mental health and diagnostic inflation, the idea that once we have supplied the public with knowledge about what constitutes depression or anxiety, once we've lowered the taboos against those diagnoses, probably that's all to the good.

14:25But there are also some people who would have thought of themselves as healthy previously who now understand themselves as struggling or mentally ill, and the effect that that has on their lives is ambiguous. Yeah, I mean, and that sort of gets to, like, I think sort of both like the benefit and the peril of the wearable phenomenon. A study in the Journal of American Medical Association found like 40 % of Americans reported wearing a wearable in 2024. And like the promise and the peril of it is mindfulness is actually pretty important. So the peril is what you just described or what I just described about my sleep.

15:00It gives you data that says, actually, you don't feel that good. Actually, your resting heart rate is kind of high. And you're sitting there thinking, but I don't feel anxious. I feel fine. And it gets you worrying. And that can obviously get into a pretty vicious circle pretty quickly. On the other hand, the benefit of wearing something like this is it can offer you data that then does the opposite, that puts you into a virtuous cycle. So like step counters, there's data for step counters that it says that it does encourage people to walk more when they're tracking because it gamifies getting exercise.

15:35I mean, the thing that I have found most sort of useful about this past week's experiment has been tracking my steps and thinking, well, I really do want to hit a certain number every day. And like, I'm going to go, you know, I'm going to go get one more walk in in order to get there. And that obviously is to the benefit. So there's a lot of stuff going on here. There's a kind of a sociological story about the sorts of people who are drawn to this? Why are people drawn to these, you know, measures of self-optimization and why are they starting to see their body in terms of data which can be extracted?

16:11Also, to what extent is that really a phenomenon of, you know, achievement culture among the well-off versus something that might be extended profitably through the rest of the population? There's that whole bucket, like the kind of cartoon, Brian Johnson, like I'm going to, you know. Do you want to say who Brian Johnson is? Brian Johnson is, I mean, it's amazing. He's a tech entrepreneur who has devoted himself to the pursuit of longevity and maybe even living forever. Yeah, never dying. The thing I care about the most is what is my heart rate before bed? Your goal in life now is to lower your heart rate.

16:48And so the way you do that, one, is you have your final meal of the day four hours before bed. And he started out as a, like, cartoon character who everybody was comfortable mocking for being so outlandishly committed to self-monitoring, self-optimization at the expense of all other human pleasure. But he's, I think, in the, like, last year sort of become, like, lovable as a completely unapologetic embodiment of something that I guess so many more of us are doing anyway. And like, we're glad that he's doing it in a cartoonish way, so uninhibited. So maybe so that we could feel better doing it in a slightly more neurotic way ourselves.

17:30I also think like the other thing about Brian Johnson is I think he's, I mean, he's definitely so like the example of this N of one experimentation that I think goes on with this, this data collection, which is, you know, I am going to track all of these things about myself and then and make these modifications, and then track what the modifications do. And the fact that there's data around it, like, it's supposed to sort of make it not anecdotal evidence, but, like, N of 1 is N of 1, right? And you can't actually, you know, tease apart causation and what is just placebo effect and all of these things when it's just your one body, right?

18:08Like, we have randomized control trials exactly because one person's experience is not enough to extrapolate to know things about the human body as a, you know, as a universal phenomenon. But even, I mean, pulling back from the end of one problem, you know, I struggle to understand how to make sense of statistics at the population level, too. Like, if I'm reading about, you know, my father's cancer or whatever, I'm talking to his doctor and his doctor's like, this person has a 20 % chance of surviving this year. I'm wondering to myself, does that 20 % describe a matter of chance? Does it describe something fundamental to his biology, which we don't understand, which we're choosing to describe by treating it as a matter of chance?

18:54And theoretically, if we could know more about this cancer and this man and his history, would we be able to produce an end-of-one assessment of what will happen with a particular cancer treatment over time? In other words, like, are we dealing with the irreducible epistemological mystery of the body? Or is it conceivable that perhaps even in the relatively near future, data, better data, better screening, better information about genes and et cetera, we could put all that into some system and actually get a reliable assessment of like, you know, when Rachel's going to die, you know, or whatever.

19:35I don't want to know. Yeah. So, okay. So two things about that. So So the first is, right, the 20 % is not a matter of chance, right? And that kind of broad statistic, in some ways, I think the problems with it are why the sort of tech folks are so bullish on the data revolution that we're talking about today. Because among other things, that 20 % chance, it's retrospective, right? It's looking at, you know, meta-analyses from studies done sometime in the last decade or the last 15 years. And it may or may not reflect what we know about Ben Sasse, right? The senator, or the ex-senator who has pancreatic cancer, who got this devastating diagnosis and was told he has months to live.

20:25And then was put on this experimental therapy and has sort of told the world that it looks as though the cancer significantly receded in his body. So, like, that's not reflected in those statistics because the science being used as treatment SAS did not exist when those statistics were derived, right? And part of what the data folks are saying is basically, like, if we collect so much data, we're just going to, like, iterate knowledge so fast. And when we give it to the robot overlords, the AI is going to read that data and it's going to see stuff that we could not possibly see. And it's going to see it so quickly.

20:58And it's going to suggest, you know, thousands of new ways to experiment on it, to, you know, to intervene, whatever. And a lot of that may lead nowhere, but some of it's going to lead somewhere. And if we just sort of like participate in that process, we're going to have this explosion of useful knowledge that will come out of collecting so much noisy data. I hear you saying that, and I find that basically persuasive on an intuitive level. I also then think about, you know, this is not the first time that we've been sold promises about what big data will do to us and improve our lives. And I think about 23andMe, which told us that we were going to not just learn about our ancestry, but also we're going to learn a lot about our health because of getting it analyzed in some centralized way.

21:47And now here we are a decade or two later. We all spit into those tubes. We got some information about where our families come from, which turns out not to be all that reliable. And the company went under. And it's like, did we actually learn anything about our health from that? And I understand that the future is big and we shouldn't always impose short timelines on promises and say, if they said it was going to happen in five years and it didn't happen by 10, that means it's a hopeless cause. But I do wonder just in a really big picture when we hear the AI leaders say casually, this is going to help us cure cancer or this will help us cure all disease.

22:20I think to myself, how should we assess that claim in a world where data has improved medical treatment but not really solved anything quite yet? So the best case scenario is that it creates a new productive tension with clinical research. So, you know, there are lots of extremely valid critiques that I share about how the clinical research enterprise is sort of broken or inadequate to our moment, too slow, driven by sort of the wrong profit motives and sort of the wrong questions and all of these things. And here along comes like this new way of being able to collect and make sense of data that is currently largely being pursued outside the clinical research framework.

23:04Like, you know, when I talk about end of one experiments, I mean, like there are, you know, thousands, maybe millions of people who are tracking things about themselves and then making changes to their lifestyle and then learning things theoretically about their own health that in aggregate might be really useful for everyone to know. But the problem with that is there are lots of different points at which things go wrong. And you mentioned like, you know, it says you're 2 % from sub-Saharan Africa. And it's like, no, you're not. You know what I mean? And that's because there's a - My wife is like sure that her dad was from India, which she definitely was not.

23:43So, right. Exactly. So like, you know, that goes to the reliability of the assessment tool, right? And Theranos. Theranos was proposed as like, you'll be able to go and prick your finger and get all of this data back. And then the problem with Theranos was the tool wasn't that good, right? But I've also had a lot of people say to me they were just too early, overpromised, and then felt forced to come to market. And if we fast forward 10 years, we're probably going to have something like Theranos that's quite useful. Yeah. And I think that that's not wrong, actually. I mean, the promise of Theranos is alive.

24:18There are companies that are getting FDA approval to basically do the 2026 version of Theranos. And there's no conceptual reason why that would not be possible. No, there's not a conceptual. I don't think that we should think of any of it as having conceptual limitations so much as questions about how you're building in rigor to figure out how you know what you know. But then in the present tense, that just makes me think, okay, so maybe the info that we get from this full body scan isn't so great. Maybe even the info that we're getting directly from our little wearables isn't so great. And maybe certain kinds of people are putting too much faith in that information and reorganizing their lives in ways that may not ultimately benefit them, may even cause them some harm.

25:05Nevertheless, we're talking about a huge amount of new information being generated, at the very least for some robots to chew through to make some hypotheses about correlations and things we may do to improve our health. And I just think, I don't know, isn't that good? I think it's potentially really exciting and good. Again, for me, it's about the rigor. There's lots of things that you can imagine being helpful to you on an individual level, like disease screening tools or other kinds of tracking. And as a physician, like, I would be so thrilled if, you know, we figured out how to detect pancreatic cancer, one of the most deadly cancers.

25:53You know, we don't have a reliable screening tool for that cancer. And if we figured one out, that would be really exciting. So it's not that I'm like either anti-scientific progress or anti-big data as a way of potentially driving hypothesis formation. But it does seem, at least the way that you're sketching it out, then theoretically concerning that so much of this self-monitoring is taking place in a sort of sociological context in which people are skeptical of doctors. They may not be actually even providing that information to any centralized source that can make use of it in a meaningful way.

26:33They're also doing a lot of stuff. I don't mean to stereotype all of Silicon Valley Twitter or whatever, but they're doing a lot of gray market peptides. They're doing biohacking of various kinds. And they are doing so thinking that they are outmaneuvering, outsmarting the slow-moving scientific establishment, not that they are serving some collective good. And that raises a couple of big questions, one of which is like, to what extent are we aggregating this data in a way that will be made useful to the population as a whole? But it's also like, who are the people who are making sense of it? Is it, you know, somebody who thinks he feels really great after having adjusted his sleep schedule in X way and is broadcasting that on social media?