Eric Topol on the Future of Healthy Aging

In this episode of The Future of Medicine, we welcome Eric Topol, MD, cardiologist, scientist, bestselling author of Super Agers, and founder of the Scripps Research Translational Institute, whose work has helped shape the field of digital medicine and the use of data, genomics, and artificial intelligence to personalize care.

Dr. Topol reflects on how medicine is shifting from treating disease to preventing it, and why extending health span—the years we live in good health—may be one of the most important goals in modern science. He explains how advances in biomarkers, wearable technology, and AI are making it possible to predict disease risk earlier and intervene before conditions like heart disease, cancer, and Alzheimer’s develop.

Dr. Topol also discusses the science behind “super agers,” people who remain physically and cognitively healthy well into older age, and what research is revealing about the roles of the immune system, inflammation, lifestyle, and emerging therapies in determining how we age.

Looking ahead, Dr. Topol shares his perspective on the future of preventive medicine, including how AI-driven prediction, organ-specific aging clocks, and new biological insights could transform healthcare from a reactive system into one focused on keeping people healthy for as long as possible.

Thank you for listening!

Call to action: If you enjoy The Future of Medicine, subscribe for more conversations with leading scientists shaping the next era of healthcare. Please rate and review the podcast to help others discover these important discussions. Share with friends and colleagues who are curious about how science becomes medicine.

Transcript

0:00:00.160,0:00:05.440
If we can just prevent one or all three of 
these age-related diseases, we have done a

0:00:05.440,0:00:10.640
most extraordinary mission for human health. 
Dr. Eric Topol is a cardiologist, scientist,

0:00:10.640,0:00:14.720
and best-selling author who helped pioneer 
digital medicine and the use of big data to

0:00:14.720,0:00:20.000
personalize care. And I realized the audience was 
the public and it changed everything. His book,

0:00:20.000,0:00:24.880
Super Agers, explores how science and technology 
can help people stay exceptionally healthy and

0:00:24.880,0:00:29.920
cognitively sharp into advanced age. We're no 
longer in the era of I don't want to know. We

0:00:29.920,0:00:34.160
want to know because there are things we can do. 
At the Scripps Translational Science Institute,

0:00:34.160,0:00:39.600
he leads efforts to bring wearables, genomics, 
and AI into everyday medicine. He joins us to

0:00:39.600,0:00:45.280
discuss how technology is reshaping patient care 
and what's on the horizon for data-driven health.

0:00:45.280,0:00:49.840
We have to get a grip on our immune system. 
Welcome to Stanford Department of Medicine's

0:00:49.840,0:00:55.680
Inside Look at the Future of Medicine.
Well, Dr. Eric Topol, welcome to the Future

0:00:55.680,0:01:00.240
of Medicine. Welcome to Stanford. Thank you for 
being here. It's great to be with you, Dr. Yu and

0:01:00.240,0:01:04.960
Ashley. Well, it's a pleasure. We've known each 
other for a long time. I've been very inspired

0:01:04.960,0:01:12.160
by you and your work, and I think your influence 
only becomes greater. You are known as maybe the

0:01:12.160,0:01:18.480
futurist in medicine — a huge influence on the way 
so many of us think about the future of medicine,

0:01:18.480,0:01:24.240
the future of healthcare. But I wanted to start 
not with the future but with the past — and really

0:01:24.240,0:01:28.320
your past — because one of the things we love to 
do on this forum is hear a little bit about how

0:01:28.320,0:01:33.520
people got to where they are. You weren't always 
a world-renowned cardiologist with hundreds of

0:01:33.520,0:01:39.120
thousands of followers and multiple bestselling 
books. Tell us about when medicine first became

0:01:39.120,0:01:42.320
a thing for you. Did you always know you were 
going to be a doctor? Where did you grow up?

0:01:42.320,0:01:45.760
Tell us a little bit about your background.
Well, thanks. And I'm flattered about the

0:01:45.760,0:01:54.640
futurist. For me, I was in college at University 
of Virginia. I was working on a thesis on the

0:01:54.640,0:02:02.480
future of gene therapy in man. This is 1975. 
That's pretty early. Yeah, gene therapy — that's

0:02:02.480,0:02:08.160
kind of the pattern with me. It always takes 
decades longer. I usually get it right but I'm

0:02:08.160,0:02:14.000
off by 10, 20, or 30 years. But everyone else is 
catching up. Anyway, while I was doing this, I

0:02:14.000,0:02:23.280
wanted to be a geneticist — that was my major. And 
I wound up working part-time on the night shift in

0:02:23.280,0:02:31.920
the UVA hospital. I got a job as a respiratory 
technician just by happenstance. I would change

0:02:31.920,0:02:40.960
the equipment for ventilators in the ICU and 
on the wards. And it hit me that these people

0:02:40.960,0:02:47.040
whose equipment I was working on — I thought they 
were going to die, and then days later they were

0:02:47.600,0:02:52.640
going to be fine. I said, wow, this medicine thing 
looks pretty good. It works. So that influenced

0:02:52.640,0:03:00.480
me to pursue it, and I had to take some extra 
courses to go pre-med during my time at UVA.

0:03:01.040,0:03:05.120
So this wasn't something you thought about 
as a kid growing up — you were headed towards

0:03:05.120,0:03:11.600
genetics. Yeah, I thought genetics was really 
the thing that interested me the most. It

0:03:11.600,0:03:19.280
was in the early stages — the whole idea of 
recombinant technology was coming about and

0:03:19.280,0:03:32.720
it really allured me. But I said, you know 
what, I've got to be practical. Wouldn't it

0:03:32.720,0:03:37.840
be great if I could be one of those doctors that 
helped these critically ill people get better?

0:03:37.840,0:03:46.240
So it was a late move within my time at UVA.
And that took you to medical school then? Yeah,

0:03:46.240,0:03:51.200
I applied to medical school and wound up 
at University of Rochester in New York.

0:03:55.120,0:04:03.280
I met my wife there. I didn't know what I 
was going to be, but I wound up at UCSF for

0:04:03.280,0:04:10.720
residency. I thought maybe I'd be a diabetologist 
or endocrinologist because my father had end-stage

0:04:10.720,0:04:16.640
diabetes with all the complications. I thought 
about being a pediatrician. And then while I was

0:04:16.640,0:04:27.120
at UCSF, I had the person who was the biggest 
influence in my career — Kanu Chatterjee. He

0:04:27.120,0:04:36.720
asked what I was going to do with my life and told 
me I had to be a cardiologist. I talked about it

0:04:36.720,0:04:41.760
with my wife and she said he's right, you should 
do that. So that's how I wound up in this field.

0:04:42.560,0:04:48.800
Pretty amazing. We've heard on this podcast so 
many stories of people being influenced, and it

0:04:48.800,0:04:51.840
seems particularly in medicine that 
single individuals often change the

0:04:51.840,0:05:01.520
course of people's lives. Sounds like that 
was the case. He really did. He recognized

0:05:01.520,0:05:09.120
something in me that I wasn't even in touch with.
So where did that take you next? After residency,

0:05:09.120,0:05:15.520
we wound up in Baltimore at Johns Hopkins for 
fellowship. Then basically I've had three jobs

0:05:15.520,0:05:21.920
since fellowship. The first was some years at 
University of Michigan where I ran the cath

0:05:21.920,0:05:29.040
lab. Then I went to be head of cardiology at 
Cleveland Clinic for 14 years, and now 19 years

0:05:29.040,0:05:34.960
at Scripps Research. The longest I've ever lived 
in one place in my life has been in San Diego.

0:05:34.960,0:05:42.720
I first got to know you — as many others did — as 
really the world's most influential cardiologist

0:05:42.720,0:05:50.640
from your position at the Cleveland Clinic. 
How do you think about that trajectory? Because

0:05:50.640,0:05:57.920
you've now spent more time in San Diego, and for 
a long time you were very focused on cardiology.

0:05:57.920,0:06:10.160
But your more recent work — the books, the content 
— is much broader. Yeah, it was cardiology,

0:06:10.160,0:06:16.560
but even in the mid-90s while in Cleveland, we 
started the first cardiovascular gene bank. Every

0:06:16.560,0:06:22.480
person who had a cardiac cath would have a sample, 
as long as they gave consent, put into a gene

0:06:22.480,0:06:32.640
bank. So I was going back to some early roots.
What happened was one of those serendipitous

0:06:32.640,0:06:39.920
experiences. I went to San Diego to start the 
first human genomics institute — there wasn't

0:06:39.920,0:06:46.880
one — even though there was a powerhouse of life 
science there. Within weeks of arrival, I went to

0:06:46.880,0:06:56.960
a program organized by Qualcomm where a guy stood 
up and said we're going to have a smartphone. This

0:06:56.960,0:07:07.040
is February 2007. We're going to have a smartphone 
connected to the internet with a camera,

0:07:07.040,0:07:17.440
and you can take pictures and send them to your 
friends. People stood up and said, "Wait a minute,

0:07:17.440,0:07:24.880
we don't need a camera on a smartphone — we have 
good point-and-click cameras. What kind of camera

0:07:24.880,0:07:32.560
is going to be on a phone? Kodak makes what we 
need." And I'm thinking, whoa. All of a sudden

0:07:32.560,0:07:44.400
I got into this whole digital medicine space. That 
was the year the iPhone was released, in November.

0:07:45.280,0:07:51.280
The timing was extraordinary because we 
were preparing a big NIH grant — it was

0:07:51.280,0:07:57.440
going to be just on genomics, but we became 
the first applicant for genomics and digital,

0:07:57.440,0:08:02.160
and it wasn't just cardiology, it was across 
the board. That was the transformation — just by

0:08:02.160,0:08:07.600
happening to be at that one meeting, and I almost 
slept through the thing. But the people raising a

0:08:07.600,0:08:12.480
ruckus about why we'd need a camera — I said, what 
do you mean why do we need it? That's a sensor.

0:08:13.120,0:08:18.880
What else could we connect to a smartphone?
And Qualcomm of all places — they make those

0:08:18.880,0:08:24.720
sensors, the electronics inside. Exactly. And I 
didn't even know when I moved to San Diego how

0:08:24.720,0:08:31.680
rich the wireless digital assets were in terms of 
the brain trust there, because the original reason

0:08:32.400,0:08:52.000
for the move was genomics. I was interviewing 
at Scripps and they were eager to get it

0:08:52.000,0:09:00.960
designated and build it, but we changed the name 
to translational because it was a broader thing.

0:09:00.960,0:09:06.640
We think of you as a very close friend of Stanford 
anyway. Things might have been so different,

0:09:06.640,0:09:14.160
but in some ways not so much — the world has moved 
in the direction it has. I think from San Diego,

0:09:14.160,0:09:19.920
which is a real hub of genomics, digital life 
sciences, biology, and biotech, you've really

0:09:19.920,0:09:29.920
been able to have a platform to influence the 
world. Was that part of your plan? The idea of

0:09:29.920,0:09:35.680
being an influencer — now they say all the kids 
want to be influencers when they grow up. You

0:09:35.680,0:09:43.520
were early. I remember one of the first times you 
visited Stanford after social media was a thing,

0:09:43.520,0:09:50.720
you asked the audience who was on Twitter, 
and even at Stanford only a minority put

0:09:50.720,0:09:59.440
their hands up. But already you were creating 
a presence and presenting information in a

0:09:59.440,0:10:06.640
way — I also think of you as the ultimate 
information gatherer and explainer, and

0:10:08.160,0:10:13.920
your Twitter feed alone is a service to the 
world. Now you have these longer-form Ground

0:10:13.920,0:10:22.000
Truths on Substack where you explain science and 
medicine to a really large group of people. Was

0:10:22.000,0:10:29.600
that something you were always headed towards?
It's really interesting you bring that up.

0:10:33.120,0:10:42.800
The idea that our audience is the public and 
not our medical microcosm only hit me with the

0:10:42.800,0:10:52.240
Vioxx debacle. This was back in 2004. Vioxx was 
withdrawn, and instead of letting it go — since

0:10:52.240,0:10:59.280
we had written the paper in JAMA three years 
prior saying we had a real problem — only three

0:10:59.280,0:11:05.920
years later they withdrew the drug, and the CEO 
of Merck claimed that was the first time they had

0:11:05.920,0:11:16.160
any signal, which was of course a lie. Anyway, I 
wrote my first op-ed. This is 2004 — 21 years ago.

0:11:16.160,0:11:26.000
And I realized the audience was the public, and 
it changed everything. I always now encourage

0:11:26.000,0:11:32.400
everybody that your audience isn't your peers 
— it's a much broader one, because your work,

0:11:32.400,0:11:38.480
if it's going to be important, needs to include 
them and hopefully have a positive impact for

0:11:38.480,0:11:46.800
everyone. That was the beginning of the whole idea 
that writing and being on social media — that came

0:11:46.800,0:11:57.200
up beginning in 2009 — and then most recently I've 
really enjoyed Substack's Ground Truths because it

0:11:57.200,0:12:04.400
gives me a long-form chance to interview people 
like you who I think are the most interesting

0:12:05.360,0:12:16.800
people changing the future of health.
I try to get younger folks to do this,

0:12:16.800,0:12:26.640
but it's hard because to write the first essay for 
the public, or to speak — in grand rounds, I would

0:12:26.640,0:12:32.960
show those exact same slides to a lay audience. 
There's nothing I would show differently. I might

0:12:32.960,0:12:40.800
just be a little more careful about abbreviations. 
We shouldn't be using inside jargon. We have

0:12:40.800,0:12:50.880
to communicate the excitement of what we're 
doing to the broad public. I wish I had known

0:12:50.880,0:12:59.040
this 30 years before I started doing it.
It's such an important point. We've tried to train

0:12:59.040,0:13:04.480
our residents and grad students to understand 
that communicating science isn't just about

0:13:04.480,0:13:10.480
your peers — of course that's an important part 
of publication — but the people who pay the taxes

0:13:10.480,0:13:19.280
that fund our science, those are not them, mostly. 
The lay public are the ones who will actually make

0:13:19.280,0:13:27.280
a difference, and they're the ones paying for 
the science. This move towards open science and

0:13:27.280,0:13:45.280
preprints makes our science available quickly to 
everybody, but it also needs explainers. The only

0:13:45.280,0:13:51.760
real criticism I've gotten from the Super Agers 
book — which has gotten a lot of non-science-based

0:13:51.760,0:13:55.680
people to read it — is "Why didn't you 
make it even easier to understand?" I said,

0:13:55.680,0:14:03.760
I tried. My editor tried. But some things like the 
immune system and genome editing are tough areas

0:14:03.760,0:14:11.440
to distill into completely common, understandable 
language. I wish I were even better at that.

0:14:16.160,0:14:28.240
When I'm rehearsing with our residents and fellows 
about their presentations, one thing I do is point

0:14:28.240,0:14:36.400
out when slides are too inside-baseball. You've 
got to make it so anyone can understand — and

0:14:36.400,0:14:42.880
don't use your slides as a crutch. You 
should be able to communicate with passion,

0:14:42.880,0:14:54.720
and that's something it took me too long to learn.
I definitely want to get to your book because I

0:14:54.720,0:14:58.880
was really excited to read it — it's been very 
successful. But two things I wanted to clarify

0:14:58.880,0:15:07.280
for our audience: you do still see patients. You 
are an active cardiologist. And this is, to me,

0:15:07.280,0:15:12.160
the most remarkable thing — you read all of those 
papers yourself. Not everybody realizes that. I

0:15:12.160,0:15:18.160
think many assume you have a team of 50 people who 
read and post for you, but that's actually you.

0:15:18.160,0:15:23.680
The team is one. That's me. People think that, 
and it's kind of crazy, but I wouldn't want

0:15:23.680,0:15:33.600
other people to read or post for me. I've always 
felt that way. Better not to do it if you're not

0:15:33.600,0:15:39.840
going to do it yourself. Today's Grand Rounds was 
a great example — three or four times you referred

0:15:39.840,0:15:45.280
to work that came out in the last few days. So 
this is obviously a daily ritual. Do you do it

0:15:45.280,0:15:54.400
early in the morning? Yes, I get up around 5:30 
and spend a couple of hours reading almost every

0:15:54.400,0:16:05.280
day. I love ingesting what's new. I've always 
been a voracious reader — Tyler Cowen calls

0:16:05.280,0:16:13.520
it being an "inforvore," and he's even more of one 
than I am. Even when I'm on vacation or holidays,

0:16:13.520,0:16:21.120
I'm still reading. A lot of people read 
as much — they just don't share what they

0:16:21.120,0:16:28.240
read. If we did more of that, we'd all 
get smarter faster. The main difference

0:16:28.240,0:16:37.440
isn't as much the reading as the sharing.
Well, another way of sharing — and maybe the

0:16:37.440,0:16:42.800
longest form before new media — was books, and 
you've written a number of bestselling books.

0:16:42.800,0:16:49.520
Every one of the books I wrote, I've been off 
by years. Creative Destruction — I said we're

0:16:49.520,0:16:56.640
going digital in the medicine world. We're still 
not quite there, but that was like 15 years ago.

0:16:56.640,0:17:01.920
Then The Patient Will See You Now — I said it 
was all going to be democratized. We're not

0:17:01.920,0:17:07.440
there yet. Then Deep Medicine — we're going 
to be using AI. Well, we're just starting.

0:17:09.520,0:17:14.560
I've tried to write books about where we're 
headed thinking it wouldn't take as long,

0:17:14.560,0:17:18.320
and I'm always wrong about the timeline. I 
get the right idea with the wrong timeline.

0:17:18.320,0:17:22.800
I don't think you're always wrong — I think 
you're ahead. And in particular, think of this AI

0:17:22.800,0:17:27.440
revolution that's happening now. You were talking 
about the ways AI could impact care before GPT-2,

0:17:27.440,0:17:36.560
really. When I was pulling together Deep 
Medicine in 2017, I met with people like

0:17:36.560,0:17:44.080
Fei-Fei Li, Andrew Ng, and many other leaders, 
and they all thought we would have something

0:17:44.080,0:17:51.920
like ChatGPT — it just wasn't out there yet, 
it was germinating. It really was with GPT-2,

0:17:54.000,0:18:03.360
and the whole transformer architecture. It was 
widely anticipated that what happened at the end

0:18:03.360,0:18:08.960
of November 2022 was going to happen — just 
not exactly when. That it could read, edit,

0:18:09.600,0:18:20.400
write, handle images, and so on. That helped me 
feel out what our deep AI world would be like,

0:18:20.400,0:18:29.280
not knowing when it would hit. Implementation is 
different from having the potential, and we're

0:18:29.280,0:18:36.960
still early in trying to bring AI into medicine.
Obviously a huge focus here at Stanford — we have

0:18:36.960,0:18:44.560
many of the architects of the revolution here. 
I look to Stanford as a leader in this space.

0:18:44.560,0:18:57.520
I've been on Fei-Fei's AI board here and I see the 
talent pool — it's extraordinary. Particularly the

0:18:57.520,0:19:03.440
imaging AI work — leading imaging AI of anywhere 
in the country. Sharing has been part of that,

0:19:03.440,0:19:08.640
and making sure data is available to the 
world to compete and make algorithms better.

0:19:14.240,0:19:20.000
Our digital scribe technology is going well. 
Doctors really like it. Patients love it. It

0:19:20.000,0:19:25.760
brings the doctor and patient closer 
together. And bringing the LLM into the

0:19:25.760,0:19:33.520
patient's medical record — not just a secure 
version for functionality, but an LLM that

0:19:33.520,0:19:39.760
can read the patient's record including records 
from other centers where a patient has received

0:19:39.760,0:19:45.680
care — is pretty novel. It was launched here just 
a few months ago, but there are already thousands

0:19:45.680,0:19:51.680
of users. It's very transformative, given that 
digital technology had somewhat gotten in the

0:19:51.680,0:19:58.240
way of patient interaction. We've had a generation 
of fellows and residents who are good at scrolling

0:19:58.240,0:20:04.400
through screens to find the information they need, 
and now the technology can help solve a problem

0:20:04.400,0:20:12.320
that the technology itself kind of created.
Your most recent book, Super Agers — there's

0:20:12.320,0:20:20.880
so much out there on aging, and one of your 
first slides was about the supplement world.

0:20:20.880,0:20:28.880
People are desperate and believe there must 
be some untapped fountain of youth. What's

0:20:29.920,0:20:36.960
great about your book is that it's grounded 
in evidence and in our understanding over many

0:20:36.960,0:20:49.440
decades of the diseases that are killing us. How 
did you decide to write a book on Super Agers?

0:20:55.280,0:21:00.960
It's interesting. I've had a long interest in this 
— when we did the Wellderly Study, it took seven

0:21:00.960,0:21:09.280
years. The idea of what accounts for exceptional 
longevity had been a kind of mystery in my mind.

0:21:09.840,0:21:19.440
I thought we were going to nail it with the whole 
genome sequence. And we saw basically nothing. We

0:21:19.440,0:21:25.680
said, whoa, what is going on here? The lack of 
familial patterns for health span and all that.

0:21:27.680,0:21:36.160
I saw the patient Lee Roso — she was so intact 
— and I said I need to get back on this. At the

0:21:36.160,0:21:47.040
same time I was seeing her, I had patients coming 
into my clinic saying, "That's not what Dr. Attia

0:21:47.040,0:21:57.040
said." I heard this a lot — I should take one gram 
of protein per pound, I should take rapamycin,

0:21:57.040,0:22:01.920
I should get a total-body MRI. I said, what's 
going on with this guy? I had met him. He used to

0:22:01.920,0:22:08.600
live in San Diego. I read the book and said, wait 
a minute — there's a lot of things wrong here,

0:22:08.600,0:22:18.880
and some things good. And I said, we have some 
science here. Since I'm organized about papers — I

0:22:18.880,0:22:24.960
don't just read them, I have everything filed — 
I had it all ready to go once I decided I'm going

0:22:24.960,0:22:32.480
to get into the whole health span and longevity 
space. Somebody's got to do it. I pulled together

0:22:32.480,0:22:40.720
about 1,800 different citations and laid out the 
real evidence, as well as a road map for where

0:22:40.720,0:22:47.440
we can take that with prevention of disease. 
So there's practical advice as well — what

0:22:47.440,0:22:54.240
do we know now? What's the real deal on sleep, 
protein, supplements, and everything else? And

0:22:55.280,0:23:01.680
what are the breakthroughs that are occurring 
that maybe you don't know about but are worth

0:23:01.680,0:23:08.960
building on to prevent these age-related diseases? 
Because if we can just prevent one or all three

0:23:08.960,0:23:16.720
of these age-related diseases, we have done a 
most extraordinary mission for human health.

0:23:23.280,0:23:27.280
There are huge investments and companies here 
in San Diego, and up in the Bay Area, to reverse

0:23:27.280,0:23:34.080
aging and slow the aging process directly. I 
wish them well — they have billions of dollars

0:23:34.080,0:23:41.840
and wonderful investors. But that will take a long 
time and may not be safe. A lot of these things

0:23:41.840,0:23:49.040
actually induce tumors and have sequelae that 
are not so light. So I said, wait a minute, isn't

0:23:49.040,0:23:57.120
there a safer strategy? That was the inspiration 
to go after primary prevention of age-related

0:23:57.120,0:24:02.800
diseases. While I was doing the research, I didn't 
know where it was going to take me, but I started

0:24:02.800,0:24:09.280
saying, wait a minute — we have these aging 
clocks from work done here at Stanford, we've got

0:24:10.160,0:24:18.640
incredible biomarkers, advances in omics, sensors, 
and AI — particularly multimodal AI. That gave me

0:24:18.640,0:24:31.440
a whole perspective on how to put it all together.
I think longevity has become almost a

0:24:32.000,0:24:33.360
staple of popular conversation 
now — everybody understands the

0:24:33.360,0:24:37.040
idea of living a long time. But the understanding 
of living well, living healthy — putting that

0:24:37.040,0:24:43.120
in the context of your individual patient was a 
really good way to kick off the evidence trail,

0:24:44.480,0:24:49.520
because at some level we're all very aware 
that there are people out there who are living

0:24:49.520,0:24:55.840
very sharp — mentally, physically — with 
lots of connections into older age, while

0:24:55.840,0:25:03.680
others who are 40 or 50 chronologically might 
as well be 80 or 90 from an aging perspective.

0:25:04.240,0:25:11.200
The fact that we can anticipate these diseases 
20 years in advance, and find out which organ

0:25:11.200,0:25:18.400
is off-kilter in its pace of aging — when 
Tony Wyss-Coray published that paper in 2023,

0:25:19.520,0:25:25.360
I wasn't aware that work was coming, and I thought 
it was the most seminal paper I'd seen in years.

0:25:25.360,0:25:31.760
I think it's going to have a huge impact on how 
we get ahead of people's age-related diseases.

0:25:31.760,0:25:36.880
I was lucky enough to know the first author 
in his lab and be part of his committee here,

0:25:36.880,0:25:43.600
so I had a chance to see that work evolve over 
a few years. You mentioned Steve Horvath's work

0:25:43.600,0:25:50.240
in the early days and the idea of aging clocks. 
Within cardiology we thought about heart age for a

0:25:50.240,0:25:56.720
while — trying to help people understand if their 
heart is younger or older than their chronological

0:25:56.720,0:26:06.160
age. We've been slowly putting more detail on 
the idea of an aging clock. But Tony really put

0:26:06.160,0:26:12.960
significant biological detail, single individual 
tissue detail, onto that for the first time. A lot

0:26:12.960,0:26:18.240
of the questions come back to: what part is about 
measuring it and what part is about intervening?

0:26:19.600,0:26:25.120
I wonder what surprised you most as you were 
pulling all of this together. I mean, you're this

0:26:25.120,0:26:31.920
information sponge going through 1,800 papers.
For me, I hadn't been following the PTAU-217

0:26:31.920,0:26:39.440
story, and I'm actually a homozygote for APOE4. 
You showed the slide — one in four people are

0:26:39.440,0:26:44.720
heterozygotes with increased risk, but I'm one of 
the two in a hundred with significantly increased

0:26:44.720,0:26:53.040
risk. So I pay particular attention to this work. 
As a biomarker, it really does seem significant.

0:26:54.720,0:26:59.920
When I got it and it was almost zero, it was 
very reassuring — I've got a good stretch and

0:26:59.920,0:27:04.960
don't need to worry. I'll get it again in a 
number of years to see if there's any change.

0:27:04.960,0:27:11.200
But for people at high risk, it's a good test. 
It's been available in the US for over two years

0:27:11.200,0:27:20.400
and people don't know it. That's what gets me — 
all these things. Like the mammogram that tells

0:27:20.400,0:27:25.360
you you're at high risk for breast cancer in 
the next three to five years even when it looks

0:27:25.360,0:27:31.520
normal. People should have that information.
We're no longer in the era of "I don't want

0:27:31.520,0:27:38.560
to know" — the Cassandra effect. We want to 
know because there are things we can do. And

0:27:38.560,0:27:44.240
we're getting to a level of accuracy and 
timing that never would have been thought

0:27:44.240,0:27:50.880
possible. People talk about incredible 
weather forecasting accuracy coming

0:27:50.880,0:28:00.480
because of multimodal AI — but this is something 
precious to have for your health and your life.

0:28:00.480,0:28:04.560
I love that you highlighted the work by Mads 
Melbye and Bernie and the paper in Nature a

0:28:04.560,0:28:10.000
few months ago. Everyone now has a sense of 
what a language model can do — predicting

0:28:10.000,0:28:23.440
the next token. But that's also a time series of 
events, and if the tokens are not words but rather

0:28:23.440,0:28:30.080
health incidents, we should be able to bring this 
technology to really help forecast for people.

0:28:30.880,0:28:38.080
What's amazing about that is it's from millions 
of people — we don't see those patterns, we don't

0:28:38.080,0:28:46.960
have superhuman vision like AI does. And it picks 
them up. What's also fascinating is that was just

0:28:46.960,0:28:53.040
with electronic health records and some lifestyle 
data. When they added polygenic risk scores,

0:28:53.040,0:29:02.000
it took accuracy to a whole other level — and that 
didn't even include biomarkers or organ clocks. To

0:29:02.000,0:29:07.512
me that was ultimate validation that with all the 
things percolating right now, we're going to be

0:29:07.512,0:29:20.933
able to say you are or are not at high risk, and 
when. And here's how we're going to prevent it.

0:29:20.933,0:29:25.600
And the granularity of that information 
matters. People understand risk generally,

0:29:25.600,0:29:34.400
but they respond generally — "maybe I should do a 
bit more of this." But there's evidence that when

0:29:34.400,0:29:40.880
they have something very specific — this is about 
you and your risk — it becomes much more real.

0:29:41.440,0:29:46.240
See, this was the problem with polygenic 
risk scores and why they're still not widely

0:29:46.240,0:29:53.840
adopted. At the individual level, if you were 
at high risk, you didn't know when. You could

0:29:53.840,0:30:02.000
be 90 or 95 before it hits. This is different 
now. With these different layers of data and

0:30:03.280,0:30:10.560
temporal timing specificity — if the large health 
model says you're going to have a heart attack in

0:30:10.560,0:30:19.440
7.9 years — that's a whole different look. It 
might be off by a plus or minus some months,

0:30:19.440,0:30:27.680
but it's a whole different way of 
predicting risk. Not just yes or no,

0:30:27.680,0:30:34.320
but the timing is what takes it to another level.
You made the point that AI is here in the sense

0:30:34.320,0:30:39.600
that the foundation models are built and 
available. We're all in academic medical

0:30:39.600,0:30:44.880
centers working with our own data and using 
the foundation models to see what we can do.

0:30:45.680,0:30:56.880
Inference is pretty cheap these days. But one 
of our faculty made the point that if we add

0:30:56.880,0:31:02.080
all of this up and try to do it for every single 
member of society, that could be a lot of money.

0:31:03.120,0:31:06.320
Well, first of all, people choose to spend money 
on things — if you add up the amount people spend

0:31:06.320,0:31:14.000
on haircuts over their lifetime, that's a lot of 
money. Some of it is an individual responsibility

0:31:14.000,0:31:18.640
element. And decisions might be different at a 
population level from a government perspective.

0:31:20.400,0:31:24.800
The first thing is you don't do anything 
until you know someone's at high risk. The

0:31:24.800,0:31:31.040
expense of these added layers of data — which 
don't include anything particularly expensive,

0:31:32.480,0:31:41.920
mainly omics or AI algorithms — you don't even 
get to that unless someone is at definite high

0:31:41.920,0:31:54.640
risk. For most of the population, you'd start 
screening at age 40 to 50. Not the dumbed-down

0:31:54.640,0:31:59.520
approach we have for mass cancer screening, 
where everyone because they're age 40 or 50

0:31:59.520,0:32:14.480
gets a scan. Those are expensive scans. Why don't 
we define people's risk before we put them through

0:32:14.480,0:32:25.680
screening that has false positives, low yield, and 
false negatives? We can do much better than that.

0:32:25.680,0:32:32.320
The whole idea is to stop with this dumb-down age 
criteria. The idea that we define pre-probability

0:32:32.320,0:32:38.720
as just "what age are you today" doesn't make 
sense in a world where we have so much rich data.

0:32:49.040,0:32:59.760
Painting a picture of a great future is easier 
than plotting the path from here to there. I

0:32:59.760,0:33:02.880
often think about things in medicine — even 
the blood pressure cuff. We measure blood

0:33:02.880,0:33:06.880
pressure because that's a technology that 
was available a long time ago, but we still

0:33:06.880,0:33:12.880
measure blood pressure and do blood pressure 
trials. It's probably a marker for vascular

0:33:12.880,0:33:18.960
health and a few other things — we wouldn't 
invent it that way today. How do we disrupt

0:33:18.960,0:33:27.440
the past in order to move to that new world?
I think there are some things so glaringly

0:33:27.440,0:33:34.400
important that we're not attending to. We have to 
get a grip on our immune system. It's fundamental,

0:33:34.400,0:33:42.080
and it's just unfathomable to me that we're going 
to be in 2026 and we have not a single way in the

0:33:42.080,0:33:50.640
clinic to do that. Our first crack will be the 
immune system pace of aging clock. But we need

0:33:50.640,0:33:57.440
more than that. I'm hoping some really good 
work from Scott Boyd at Stanford and others

0:33:58.880,0:34:07.520
will help us get to a low-cost immunome that we'll 
get as part of a yearly checkup. We need that.

0:34:07.520,0:34:14.720
The other thing is the images — like 
the retina — are so amazingly rich,

0:34:14.720,0:34:18.800
and why aren't we using this? We're leaving all 
this information on the table. It's literally a

0:34:18.800,0:34:24.320
window into your blood vessels — the only place 
you can see them. You can tell about the heart,

0:34:24.320,0:34:33.760
the heart blood vessels, the kidney, the liver — 
risk of everything. And it's an inexpensive test.

0:34:33.760,0:34:40.720
It should be part of our annual eye exam as we 
get older, and not just getting the picture but

0:34:40.720,0:34:49.280
getting AI analytics. We'll get there eventually.
A lot of patients today are taking their labs and

0:34:49.280,0:34:54.400
putting them into ChatGPT or Gemini or whatever, 
and it's telling them things their doctor didn't

0:34:54.400,0:35:01.920
tell them — explaining things. We're going to see 
a lot more of that. But I think the US is not well

0:35:01.920,0:35:10.240
positioned for a large-scale preventive medicine 
campaign because of our misaligned incentives.

0:35:10.240,0:35:16.480
It's much more likely that other countries 
— because they care for patients at a true

0:35:16.480,0:35:24.640
population level, not with insurance companies 
and all the ridiculous expense we have here — are

0:35:24.640,0:35:34.480
going to invest and exploit AI for prevention. 
That small investment will pay huge dividends for

0:35:34.480,0:35:43.280
reducing the burden of disease and the economic 
hit. Whereas here, I don't yet see the chance — we

0:35:43.280,0:35:47.680
don't have that closed system where there's a 
real incentive to put the work in early to make

0:35:47.680,0:35:53.120
the savings later. So I think we're going to 
see a lot more super agers outside the US as

0:35:53.120,0:35:57.840
a result of preventing these age-related 
diseases. I'm obviously very disappointed

0:35:57.840,0:36:02.800
about that because we have the potential 
here, but it needs a lot of restructuring.

0:36:09.760,0:36:13.600
Talking of other systems — I know you spent time 
in the UK looking at the National Health Service,

0:36:13.600,0:36:20.000
and the Topol Report has been pretty 
influential in their policy making over

0:36:20.000,0:36:24.880
the last few years. That's a very different 
— and very closed — system. What were some

0:36:24.880,0:36:35.600
of the real takeaways from that process?
I think the UK is likely to be the one to

0:36:35.600,0:36:48.480
do this first. I was amazed at their aspiration to 
use AI and change the workforce to provide better

0:36:48.480,0:36:58.000
care. I had an incredible experience working with 
almost 50 people on a team to pull together — I

0:36:58.000,0:37:04.640
didn't want it named after me, but they did that 
— this NHS review. We came up with a lot of ideas,

0:37:05.280,0:37:11.920
and I think they're on it now. They're obviously 
going through a lot of economic hardships

0:37:11.920,0:37:21.200
for the NHS that weren't all there when I was 
involved, but they're still committed. The idea

0:37:22.240,0:37:28.480
is: how do we prevent these major age-related 
diseases now that we have the potential to do

0:37:28.480,0:37:37.360
it? The whole idea is to do a randomized trial 
— like we're going to do in Alzheimer's starting

0:37:37.360,0:37:44.800
in January — and then get to a point where we 
have interventions, call them preventions, and

0:37:44.800,0:37:53.120
then export them to the UK to go to grand scale.
One of the big things about the metrics of aging

0:37:53.120,0:37:59.600
we have is that we can use those as surrogates to 
show we're making a difference — you can't wait 15

0:37:59.600,0:38:05.840
years like the UK Biobank's 17-year follow-up. 
We can't wait that long. If something really

0:38:05.840,0:38:14.400
changes the trajectory of a major disease, 
we'll know it by these clocks and markers.

0:38:14.400,0:38:21.600
You were getting at this with your question about 
what surprised me pulling those 1,800 papers

0:38:22.400,0:39:04.080
together. When the obesity story was cracked with 
the GLP-1 drugs, it made me think — we'd been

0:39:04.080,0:39:09.600
working on that for decades and most people were 
going to give up. And now here we have something,

0:39:09.600,0:39:18.000
even though it looks like a forever drug, with 
huge impact. The cost will keep coming down,

0:39:18.000,0:39:22.000
it'll be made into pills, and so on. 
So we have a precedent for something

0:39:24.080,0:39:29.680
profoundly difficult being solved. Aging is 
even more difficult than obesity. But with all

0:39:29.680,0:39:37.200
the investments and some really good science, I 
changed my mind from "impossible" to "possible."

0:39:39.040,0:39:46.000
There are so many different shots on goal.
The question is, if this is really expensive

0:39:46.000,0:39:52.080
and it slows aging by a year or two — it would 
be kind of like some of our cancer drugs today

0:39:52.080,0:39:56.800
where you pay hundreds of thousands of dollars 
and get three more months. That's not going to

0:39:56.800,0:40:03.280
be the world-changing impact we're seeing in 
obesity intervention. It has to be a really big

0:40:03.280,0:40:09.440
impact — a Benjamin Button kind of thing. I don't 
know if we'll see that. Plus I am worried that

0:40:09.440,0:40:18.560
when you start with stem cell factors, senolytics, 
or these other strategies, this isn't kids' stuff.

0:40:19.600,0:40:27.200
Instead of just getting the markers showing 
clocks have reversed — great — what about the

0:40:27.200,0:40:34.400
downsides? We can't just use the markers for 
these potent interventions to reverse aging.

0:40:34.400,0:40:38.560
We have to get long-term follow-up.
Some people jump the gun and say,

0:40:38.560,0:40:44.720
"If I can get three years younger across the board 
on my clocks and body-wide aging, I want it." But

0:40:45.440,0:40:51.840
they may be paying a price for that. And we have 
lots of people on rapamycin or metformin. One of

0:40:51.840,0:41:01.040
the other issues in the book is people asking what 
dose of rapamycin they should take. We have no way

0:41:01.040,0:41:07.840
to measure the immune system. These advocates 
— Attia, Bryan Johnson, and so many others,

0:41:09.120,0:41:15.680
Huberman — telling people they should take 
rapamycin, all at different doses. And now

0:41:15.680,0:41:20.880
they're saying don't take it because I get 
infections, after years of telling people

0:41:20.880,0:41:27.600
they should take it. We have no way to check the 
immune system. Somebody could take a low dose and

0:41:27.600,0:41:34.400
their immune system could be shut down. It would 
be one thing if we could measure and titrate so

0:41:34.400,0:41:41.120
that you don't disrupt immune protection. 
But to play with that kind of firepower

0:41:41.120,0:41:49.200
without measurement — I'm amazed at that.
So what is your prescription for all of us if

0:41:49.200,0:41:58.640
we want to be super agers? The lifestyle factors 
— there are like 20 of them — are fundamental,

0:41:58.640,0:42:04.480
but we know we can't get everybody to do all 
of this. We're talking diet, exercise, sleep.

0:42:14.880,0:42:22.480
Exercise is number one, with the biggest impact — 
not just aerobic, but also resistance and balance.

0:42:26.320,0:42:33.040
The research has been so compelling. Sleep health 
I changed too — I wasn't paying enough attention

0:42:33.040,0:42:39.440
to that. I was a poor quality sleeper, not just 
in terms of quantity. I really worked on that.

0:42:39.440,0:42:48.800
And of course, having an anti-inflammatory diet, 
being at the right weight. Then you've got social

0:42:49.920,0:43:02.400
connections, purpose, being out in nature. The 
data for being out in nature is quite amazing.

0:43:03.680,0:43:08.960
We're lucky in California — we have a 
better shot than people in the tundra

0:43:08.960,0:43:16.640
zone where I lived in Cleveland for many years.
There are many more factors beyond that. The point

0:43:16.640,0:43:23.680
is that's the easiest path if everybody could do 
it. But in order to get people to be super agers,

0:43:23.680,0:43:32.000
we have to go beyond that. I know too many 
patients and friends who are very conscientious

0:43:32.000,0:43:38.880
but still get these bad diseases. So what can we 
do to prevent them? That's where we need immune

0:43:38.880,0:43:45.040
system modulators. We need things that block 
inflammation in the brain and the body — that's

0:43:45.040,0:43:55.520
where the GLP-1 drugs look very good. And other 
things specific for the organ age-related disease

0:43:55.520,0:44:02.240
at risk in that individual. We're going to get 
there. I'm confident we will make a huge dent

0:44:02.240,0:44:07.200
in age-related diseases going forward.
And this, I think, is going to be the

0:44:07.200,0:44:13.280
singular most important contribution of AI 
in medicine — preventing diseases. People

0:44:13.280,0:44:17.840
are talking about drug discovery, and drug 
discovery is going to help prevent diseases,

0:44:17.840,0:44:25.600
not just treat them. All the other things 
we're doing with AI in medicine — accuracy,

0:44:25.600,0:44:31.680
increasing humanistic interactions — they're 
great, but this one I think is going to wind up,

0:44:31.680,0:44:35.680
years from now, being the biggest.
Well, apart from anything else, Eric, we're

0:44:35.680,0:44:40.480
very happy that you're focused on super aging so 
we can have another 20 or 30 years of you leading

0:44:40.480,0:44:44.960
us into the future. I hope that comes about. 
But thank you — really appreciate it. Thanks

0:44:44.960,0:44:50.880
for joining us here. It was so much fun. Thanks.
The preceding program is copyrighted by the

0:44:50.880,0:44:59.120
Board of Trustees of the Leland Stanford Jr. 
University. Please visit us at med.stanford.edu.