Measuring the Pace of Aging


Daniel Belsky

Associate Professor of Epidemiology, Robert N. Butler Columbia Aging Center, Columbia University

Dr. Belsky is an Associate Professor of Epidemiology in the Robert N. Butler Columbia Aging Center at Columbia University. His studies seek to understand how genes and environments combine to shape health across the life course. The goal of Dan’s work is to reduce social inequalities in aging outcomes in the US and elsewhere. With collaborators Terrie Moffitt and Avshalom Caspi he originated the Pace of Aging method to quantify the aging process from longitudinal analysis of human physiology and recently translated this method into a DNA-methylation blood test that can be implemented from a single time point of data collection. He is principal investigator of NIH-funded projects to test how caloric restriction may slow or reverse aging-related changes to the genome (CALERIE), to understand long-term impacts of in-utero famine exposure on biological aging (Dutch Hunger Winter Family Study, with Bertie Lumey), and to test the potential of anti-poverty policy intervention to slow biological aging (MyGoals for Healthy Aging, with Peter Muennig). Dan’s work has received international attention, including by the Wall Street Journal, Washington Post, and Guardian newspapers, and appeared in outlets including PNAS, Nature Human Behaviour, Nature Aging, the JAMA journals, Lancet Respiratory Medicine, and top journals in epidemiology and gerontology. Since 2020, he has been named an ISI highly-cited researcher.


Episode Transcript

I think that we have it within our power to change the environments we live in, in ways that can make profound differences in our health span. And I think we’ve already changed environments in ways that have made profound differences in our health span. We see it around us every day. The challenge we face now is how do we expand access to all those extra years of healthy life that the most fortunate among us are able to enjoy? 

Aging like gravity, it pulls on each of us. Why do some of us age without illness? How do our bodies and minds experience aging at the silver and molecular level? What’s the future of aging in our society? And maybe most importantly, what can we do about it today? My name is Gordon Lithgow, and here at the Buck Institute in California, my colleagues and I are searching for and actually finding answers to these questions and many more. On this podcast, we discuss and discover the future of aging with some of the brightest scientific stars on the planet. We’re not getting any younger, yet!

 Hi everyone. Welcome to the show. Today with me is Daniel Belsky. Daniel is an associate professor of epidemiology in the Robert M.  Butler Columbia Aging Center. We’re going to be talking about the intersection between public health, population, and behavioral sciences. And this really is a brave new frontier where we see geroscience beginning to make an impact on the thinking of people who have been thinking about public health issues. And it all starts really with the way in which we measure aging. And I’m really keen to talk to Daniel about, his measures and the pace of aging, along with other clock-like measures. This is going to be a really interesting conversation. 

 I ‘m absolutely delighted to have Daniel Belsky here today. I’ll be frank- I’m jealous! Because you get to follow, it seems like you just get to follow whatever you’re interested in. You know, you’re not, you’re not tied to a particular branch of biology or epidemiology. Tell me what your average day is.

Read more close

Daniel: [Laughs] I don’t know that I have too many average days, to be honest.

Gordon: [Laughs]

Daniel: but-but in a given day, you know, I-I’ll probably work on you know, two to three different projects m-meeting with-with different students and collaborators. And they can really span as you suggest, a fairly broad spectrum of work. So we have a project focused on the long term outcomes of individuals who were exposed to a famine during gestation.

Gordon: Yeah.

Daniel: so the-the there was a very famous famine in Europe in the middle of the 20th century called the Dutch Hunger Winter in which a German blockade of the Western Netherlands caused a flash famine that lasted for about four or five months, before the Allied liberation of the Netherlands in the spring of 1945. And the children who were in utero during that period grew up to develop increased risk for obesity and-and signs of risk for metabolic disease. So we were interested in whether that early insult might, in fact, contribute to an overall acceleration of biological processes of aging. So-so that’s one project —

Gordon: Yes.

Daniel: — looking at the very beginnings of life.

Gordon: Yes.

Daniel: We have other work interested in how the socioeconomic conditions, the neighborhood characteristics, the parental education, the household wealth that children grow up with and then develop for themselves as they grow into adults shapes trajectories of aging from midlife through the later part of the lifespan. And we have projects looking at ways in which intervention might slow the aging process.

Gordon: Mm-hmm.

Daniel: So we work with the calorie trial which is a first long term trial of caloric restriction in healthy, non-obese humans that we’ve generated a range of genomic data for to understand how this intervention which we know slows aging in worms, flies and mice may translate to humans. And our interest in caloric restriction is not so much as an intervention whose biology particularly fascinates us —

Gordon: Yeah.

Daniel: — or that we think is scalable from a public health perspective. I — in fact, I think quite the opposite.

Gordon: Mm-hmm.

Daniel: But rather, as a benchmark for what may be achievable with alternative modes of intervention that would be more scalable in the general population. And we think about things, you know; here at the Buck people are interested in molecules we might deliver into people —

Gordon: Yeah.

Daniel: — to change their pace of aging. But back at the School of Public Health in New York, where I work, we’re thinking about how we can change the environments that people live in, to achieve the same kinds of effects.

Gordon: I’m particularly excited to have you here because you’re a won — a wonderful communicator of the science that you do, and-and I need you because my head is spinning with clocks: there’s methylation. There’s proteomics. There’s transcriptomics- tell me your definition.

Daniel: We all learn the word clock as a reference to some kind of algorithm-based measurement of aging from Steve Horvath in-in 2013 when he published his seminal paper showing us that from just a drop of blood or, you know, frankly a little scraping of skin, or a cheek swab, he could tell us how old the person who had given us the sample was. And that was fairly miraculous frankly, because in the case of blood, which is what we spend most of our time studying, the cells have only been around for at most, a few weeks, and somehow, they know when we were born.

Gordon: Were you skeptical when you first read that paper?

Daniel: I think that I was kind of blown away by the volume of data that Steve marshaled —

Gordon: Yeah.

Daniel: — to prove his point. By the time I encountered the science, Steve had been pushing this boulder up a very steep hill for a very long time. And when the paper was finally published in-in genome biology he had accumulated really, an extraordinary array of validating observations to establish that the methylome really did know how old the organism was. And so, it was you know, to me, it was kind of a shock and awe presentation of data and I was fully persuaded that you could predict how old an organism was from the methylation state of their DNA. The interesting question, of course, is-is, “How do we interpret the errors in that prediction?”

Gordon: Just explain the word methylome, if you will.

Daniel: Sure… So Steve worked with a molecule that up to that time had not received a great deal of attention in biomedical research. There are many different epigenetic alterations of the genome. These are chemical changes that annotate our DNA sequence and regulate the way it is expressed. A senior colleague of mine, Andrea Baccarelli, uses the metaphor of the notes a conductor makes on a-a —

Gordon: Hmm.

Daniel: — musical score —

Gordon: Nice, nice, yeah.

Daniel: – to help interpret the way the music is read.So DNA methylation are among those family of epigenetic marks. They’re methyl groups that bond to particular nucleotide pairs o-on the genome. And they recruit chromatin, and I was originally taught that they acted to regulate the expression of genes. Particularly, when they bond at a particular position relative to the protein coding region of the sequence. So we’re —

Gordon: Mm-hmm.

Daniel: — we’re fairly deep in the weeds, and it’s probably not responsible for me to go much further not being a card-carrying genome biologist. But what I’ve learned over the years is that that canonical model of the way the methylome works may, in fact, be wrong and, in fact, what we study when we’re measuring these chemical tags that are strewn across the genome is something that we don’t fully understand.

Gordon: Hmm.

Daniel: These are signals of an otherwise unobserved biological process from which we are inferring the progression of this family of molecular changes that we call aging.

Gordon: Yeah.

Daniel: And so, you and your colleagues here —

Gordon: Yeah.

Daniel: — at Buck have elucidated many of these so called hallmarks or pillars of aging and we’re not directly observing those when we quantify DNA methylation across the genome. Rather, we’re examining a molecular substrate that we think carries a signature of those changes. And that’s what we’re interested in interpreting. Now we can use relatively straightforward machine learning methods. These days, we might call them AI if we want to sound fancy to form algorithms that will predict some value. And —

Gordon: Yes.

Daniel: –and so, I think that the way the word clock is being used now in biomedical research is as a blanket term to refer to any algorithm-based biomarker integrating signal —

Gordon: Okay.

Daniel: — from several, to a dozen, to hundreds of measurements, nto a single number that reads out on the functional integrity of an organ –

Gordon: Yeah.

Daniel: — or in-in the case of these aging clocks the biological age or, in our work, pace of aging of an individual.

Gordon: Okay, so that would apply to methylome. It applied to the proteome. Changes in protein levels in different tissues during aging, I guess. And —

Daniel: The metabolome —

Gordon: Yes.

Daniel: — the transcriptome —

Gordon: Yep.

Daniel: — and-and other “omes” that I haven’t thought of and maybe haven’t even been invented yet.

10:13:53 Gordon: Yeah.

Daniel: But broadly, we can imagine or-or maybe I should s-step back for a second. I-I mean, I think — I think what’s important in thinking about clocks, is that these are largely a data-driven methodology for measuring in our case, the aging process and other cases perhaps the health or function of an organ. And that’s a transition away from an approach that previously was dominant which was much more of a hypothesis-driven approach where model systems experiments fed forward candidate mechanisms which we would then try to assay directly.

Gordon: Mm-hmm.

Daniel: And those assays were our initial set of biomarkers of-of aging —

Gordon: Yeah.

Daniel: — for example.

Gordon: Yeah.

Daniel: And they had value. We made some progress but there were a lot of things they couldn’t do. And it turned out our intuition for what was biologically interesting and that could also be measured especially in free-living humans wasn’t always the best. and that —

Gordon: Yeah.

Daniel: — that led us toward these data-driven approaches which became available to us as we developed assay technologies that could comprehensively survey variation across the genome. Across the proteome. Across the metabolome, et cetera. And so, now these clocks are developed by doing some of these comprehensive assays, and then data-mining those datasets to derive these prediction algorithms.

Gordon: Right.

Daniel: And a lot of what I spend my time thinking about and what we work on is what the labels should be in that data-mining exercise. So if you think about clock-making. Clock-making essentially consists of conducting comprehensive measurement of some molecular substrate or perhaps several. And then fitting that universe of data to some kind of phenotype. And we call that some kind of phenotype in machine learning language that’s the ‘label’. And the initial label used in biological aging clock work was the age of the participant who had given their —

Gordon: Chronological age, yeah.

Daniel: — their biospecimen for analysis. Chronological age.

Gordon: Yeah.

Daniel: That’s right.

Gordon: Yeah.

Daniel: These were algorithms that predicted how long it had been since the person had been born to the time when they gave us their DNA.

Gordon: The other side of this activity are the biologists and the clinicians collecting functional data on aging phenotypes or disease perhaps. In other words, this relies on two parts. It relies on the algorithms and everything you just talked about, but it also relies on the quality of the assessments of health at the other end. How — where do you think we are with that? Are you comfortable with the datasets that are out there that people are using to derive clocks, and then, you know, make predictions from clocks?

Daniel: This is — excellent question. There are two dimensions here. and one of them is again, what’s the label? What are we measuring, as our surrogate for this underlying process of aging that we’re-we’re going to point our genome at? Our methylome. Our proteome. And the other is who are we measuring it in? And I think that on both fronts there’s a great deal of room for improvement but we’re making progress. And so, on the-the what do we measure front we have moved beyond chronological age.

Gordon: Yeah.

Daniel: First, to survival time so essentially looking in the opposite direction from the genome to the end of life instead of back toward the beginning. But then, you know, for example, our work is interested in charting changes that occur within people’s bodies over time. And setting as the label in our analysis, the rate at which those changes are occurring. So we’re interested in our case in the rate of physiological deterioration occurring across a decade or two decades of adult life. And there are other phenotypes emerging in this work that involve compositing analytes reflecting the function of specific organ systems or perhaps — this is work in the pipeline — response to particular stressors or challenges.

Gordon: Mm-hmm.

Daniel: So I think that we are — we’ve made a lot of progress. We have some ways to go, and it will be hard to know what the — what the perfect label is.

Gordon: Yeah.

Daniel: And there may be no perfect label.

Gordon: Yeah.

Daniel: We may need a set of different labels and-and multiple algorithms to be able to create a comprehensive picture of-of human aging.

Gordon: But would that include a frailty index, for example, which tries to accumulate data from different types of measures?

Daniel: Yeah, so the frailty index is an example of a composite phenotype of human health. It is most useful in quantifying individual differences. Differences between people in aging at the end of the lifespan. And so, we think of aging as a process ongoing from the very beginnings of life. That’s something we learned from the work of Vadim Gladyshev in his lab. And what we’re after are metrics that can illuminate differences between people in the pace and progress of that lifelong process. Really, when we measure them in childhood and adolescence. In early adulthood. In later adulthood. And for that reason frailty has not been the most appealing label to us. Nevertheless, a key application of geroprotective therapies of these new interventions to slow —

Gordon: Mm-hmm.

Daniel: — aging will be in preserving health in older people. And in that context algorithms trained to predict frailty can be exceptionally useful. And I expect that we’ll see a number of them in the years ahead.

Gordon: Okay, great.

Daniel: Um —

Gordon: That —

Daniel: But before we fully transition I just want to come back to that second dimension which is who we measure.

Gordon: Yes.

Daniel: And to date, most of the clock-making has been done with populations of convenience. What we call the usual suspects in biomedical research. So these are reasonably well-off, reasonably well-educated people who live near academic medical centers and are bio-curious. They’re interested in their own health. They’re interested in what can be learned about it.
These people are great. We count on them, but they don’t represent the universe of patients that we’re seeking to serve with the tools that we develop. And so, there’s been a push to diversify the samples that we use to develop and validate these clocks to include more representative samples of more populations. So not just people who live near academic medical centers, but people scattered across rural and urban areas around the country. People from different racial and ethnic backgrounds. People from different socioeconomic backgrounds.

Gordon: Yeah.

Daniel: And that was the — one of the-the really exciting opportunities we had in the work I did with the Dunedin Longitudinal Study where a population-based cohort had been established from a birth register. So all of the babies born in a particular hospital, in a particular city, in a particular year. And had been followed forward across when-when we did our work for decades. And now the cohort is pushing into the fifth decade of its follow-up.

Gordon: And is that a diverse population that-that — because it’s centered in a single hospital? That must be in a particular geographical —

Daniel: That’s absolutely right, so —

Gordon: — socioeconomic niche.

Daniel: So it is a particular geographic setting, and it is a particular, in fact from an ancestry perspective genetic setting. The Dunedin Longitudinal Study was initiated in Dunedin, New Zealand, 1972 to 1973. And the population of Dunedin, New Zealand is overwhelmingly descended from immigrants from Scotland. So the sample is almost entirely white, and there is some indigenous representation in the cohort but their samples were actually not included in our analyses because their blood cannot leave the island of New Zealand.

Gordon: Aha.

Daniel: so-so we’re-we’re not —

Gordon: Yeah.

Daniel: — working with, in the case of the Dunedin Study a race or ethnically diverse sample, but they are socioeconomically diverse. Representing the full spectrum from kids who grow up in conditions that are — that are — that look like poverty–

Gordon: Yeah.

Daniel: — to those who are growing up very well-off. and because of the extraordinary efforts of the investigators, first Phil Silva and his colleagues at the University of Otago. Later, Richie Poulton. And on our side of the ocean Terrie Moffit and her partner, Avshalom Caspi the cohort has retained over 95 percent of the surviving participants —

Gordon: Fantastic.

Daniel: — across more than a dozen waves of follow-up allowing us to observe the aging process in a relatively unbiased manner in this particular population.

Gordon: Yes. Well I’m encouraged to hear people of Scottish descent are being studied. That’s very valuable.

Daniel: [Laughs]

Gordon: So it’s a-a very interesting population, I guess, from-from my perspective. Certainly has all the characteristics of a population with highly diverse social, socioeconomic status. As you know, life span, life expectancy within the city of Glasgow is everything from the best in the world to almost Somalia levels. So. So it’s a very interesting population, I guess, from my perspective. Now Daniel, you used the word pace at some point there. And you’ve published on this idea of pace. Just, just can you just outline a definition for that and then also contrast it to maybe some of the other things like healthspan.

Daniel: Sure, so I think that when we started doing this work I was coming out of a postdoc at the Duke Aging Center. And with Terrie Moffit and Avshalom Caspi we were trying to figure out how we could advance the measurement of aging. We were interested in setting up an end point for the ways in which early life exposures shaped the aging process, but there was no established measure of aging so we were trying to come up with something that would be valid. And the idea at the core of that effort was that aging was about change. That we had measures of disease that could differentiate people who-who were in good shape from people who were in bad shape. But what we were ultimately interested in understanding was how change unfolded in peoples’ bodies and whether during young adulthood, early midlife before the onset of chronic disease, it was possible to detect differences between people in their aging process. So, what ultimately became the pace of aging was an effort to observe within individual lives first, over a period of a dozen years and-and in the follow-up study 20 years, how different organ systems and blood analytes changed across age 26 to age 32, to age 38 in-in our first study, and then on to age 45 in the follow-up. We were interested not just in determining whether we could detect change in peoples’ bodies, but whether people varied in how fast those changes unfolded. And so, what we ended up calling the pace of aging were what was the dimension of individual differences in how fast changes occurred across the set of organ systems we were able to observe. So we, because of the vision of the Dunedin Study investigators Richie Poulton and his team at the University of Otago, Terrie Moffit and her team at Duke University, had accumulated parallel measurements of the cardiovascular system. The pulmonary system. The liver or the kidneys. The lungs. Even the gums, because of the routine dental examinations, across these repeated measurements of the study participants who every five or six years were brought into the University of Otago and received a full day of measurement activities. And, in fact you know, one of the ways in which the study is able to get such a high participation rate is that in the cases where people are living abroad they offer them a plane ticket to come back. And in the cases where they can’t come back they send a research worker to go find them.

Gordon: Wow.

Daniel: And they’ve even gone to visit people who have been incarcerated, so it’s an enormous effort to-to keep everyone involved in the study. So what we define as the pace of aging is the rate at which a person’s body is changing relative to the norm for their sex, we did this separately for men and women, and the source population. And we take the Dunedin Study as representative of the source population in Dunedin, New Zealand. So unlike the-the other clocks, either the ones that predict your chronological age or the ones that predict your survival time which really function like an odometer on a car.

Gordon: Mm-hmm.

Daniel: They tell you distance traveled.

Gordon: Mm-hmm.

Daniel: “How much aging have I done so far?” Dunedin — the pace of aging attempts to function more like a speedometer for the aging process. “How fast am I traveling?” and the goal there was partly in identifying something that would be sensitive to exposure histories in advance of the onset of chronic disease. And also, something that could be malleable in response to interventions that would modulate the rate of aging.

Gordon: Thank you, that’s really helpful. Again, you’ve alluded to this need to think about early life, the biologists are all running around finding new interventions all the time actually. At least in-in animal models in the laboratory. And they’ve freezers full of compounds that extend lifespan and do interesting things in these systems. One of the bottlenecks, I guess, is biomarkers. A second bottleneck, I think, is-is bringing those compounds forward into some sort of meaningful experiments that then you can go off and impress a clinician with. But talk about early life because it gets into a difficult conversation quite quickly. About the idea of potentially giving agents that we hope are going to intervene in aging and extend health span, and compress morbidity, and all that good stuff. But we’re going to potentially be asking young, healthy people to indulge in those kind of interventions?

Daniel: Well, so I think what we mean by young people could be different. So I want to be clear. I think that we understand the aging process as o-one that-that unfolds in parallel to development from the very beginnings of life. The accumulation of molecular damage and the breakdown in the integrity and resilience capacity of cells, tissues and organ systems is really competing with the organizing force of de — of development as we assemble the organism. And then ultimately, you know, we see — we see the other side of it during the aging process. The way in which I think about the utility of an understanding of biological aging and the measurement of it in children is in service of understanding how the environments that we cultivate for young people shape their aging trajectories and can contribute to building healthy lifespan. So not at all drug interventions.

Gordon: Yeah.

Daniel: Now, to the question of whether young and midlife adults might benefit from these interventions I think that’s an open question. But it is the case that relatively young and healthy people are the ones likely best able to sustain any adverse effects of geroprotective interventions. And so, they may provide a pool of volunteers for early-stage translational work that might not be safe for individuals who are older and more frail. Now I understand that that could be complicated from a regulatory perspective and that’s in some ways a bit beyond my area of expertise. But our interest in devising metrics that could read out on the efficacy of geroprotective therapies on our ability to slow aging in people in young and midlife adults was that these are the people for whom prevention of aging-related disease is still possible. So one of the things that we know from public health is that preventing or delaying the onset of disease is an easier lift achievable for more people than delivering a cure.

Gordon: Mm-hmm.

Daniel: And so, ideally, our therapies can support the preservation of health for longer rather than simply try to recover it once it’s been compromised by a disease process. And that’s a different approach than-than we get in standard biomedical research or-or clinical work. But I think it’s the one that we ultimately need if our goal is to engineer a population that lives more years of life in better health.

Gordon: As you know, there are widespread gender differences in response to interventions especially in mouse models. Generally, they work in one sex and not the other. Are you seeing anything across this wide view that you have of public health? Are you seeing those gender differences? Are women aging in a different way than men? And do we need to think of interventions that are sex-specific?

Daniel: In the work that we’ve done so far, we have seen very little suggesting that the kinds of broad environmental and social context that we study, affect aging in different ways in men and women. And in the caloric restriction trial we saw parallel responses in men and women although there are other measurements that have been taken from that trial that have some sex differences in response. And obviously, as you say, in the animal models there are sex differences everywhere. From a descriptive standpoint many studies find that men appear to be aging more rapidly than women across the middle and into the later part of the lifespan. And that’s consistent with the shorter healthy lifespan of men, but perhaps inconsistent with the observation that in later life women tend to accumulate more morbidity than men do. And from a-an epidemiology or-or biodemography standpoint we tend to resolve that paradox by reminding ourselves that one of the reasons there may be more sick women in the population of older adults than sick men is that the sick men don’t survive.

Gordon: Yeah.

Daniel: and that-that problem of survival bias is a big deal in epidemiology, and I think remains a major challenge in the way we study the biology of aging. Particularly, when we think about studies that use comparisons of older to younger people —

Gordon: Mm-hmm.

Daniel: — to define what the biology of aging is. And the reason for that is that older people, let’s say, people 70, 75, 80, these people are already exceptional survivors!

Gordon: Yeah.

Daniel: They’ve outlived half the babies born in the same year that they had. And so, the biological differences we observe between this population of exceptional survivors and the population of average survivors that we observed decades earlier in the lifespan —

Gordon: Yeah.

Daniel: — could be the depredations of aging, they could be the accumulation of biological damage, or they could be signs of the very resilience that’s allowed these people to survive this long. And so, one of the challenges we have in devising aging biomarkers is how to sort that out. And it’s one of the reasons that we’ve continued to pursue these pace of aging biomarkers that try to set apart differences between people in how old they are from differences in how their bodies are changing over time.

Gordon: So before we move on, Daniel, obviously one of the most famous interventions of all is caloric restriction which feeds into quite a lot of your studies and your thinking. Just give us a critique of wh-where caloric restriction is at in the lab, and then in people.

Daniel: Well, I don’t know if this is a critique necessarily. You know, I’ve been awed by the evidence that’s been accumulated over really, a century of research on the ways in which calorie restriction affects that aging process. It’s a tremendously valuable model for understanding the malleability of aging biology and also, the mechanisms through which that-that malleability is achieved. from a public health standpoint though we don’t think of it as something that is highly scalable. And then as we think about it from the practical standpoint of running trials, moving from animals to humans we face the challenge that when you want to calorically restrict a worm or a mouse you just feed them less.

Gordon: Yeah.

Daniel: And you get precisely the dose you want, but when you’re intervening with humans who, you know, kind of leave the lab, and go home, and do whatever they want —

Gordon: Yeah.

Daniel: — it’s a much harder challenge.

Gordon: Yeah.

Daniel: And so, we have extraordinary heterogeneity in the adherence of individuals to interventions even under the best possible conditions. So I was fortunate to work with the calorie trial where they had teams of nutritionists and dieticians, psychologists working with participants to help keep them on the diets that they were on. They were shifting diets when they weren’t working out for people, and these were participants who were extraordinarily motivated to get this done. And it still proved incredibly challenging for them, so it’s a different enterprise when you move out of the system you fully control to one where you’re just asking.

Gordon: Yeah, humans are difficult. I know that-that you know, there was one trial looking at liquid intake in the elderly. And they were stunned that these elderly people, you know, in nursing homes were drinking gallons, and gallons, and gallons of water which was the prescribed liquid. Until someone pointed out that what they’re really doing is chasing the high from the caffeine they normally got from tea, [laughs],which is their normal drink… so yeah, humans are difficult to control.


Gordon: You are very interested in the non-genetic components and-and some estimates of hereditability of — for life expectancy in humans are very low- under 10 percent. we’ll get your take on that to begin with. But then let’s get into the idea of the exposome and how on earth we’re going to study this.

Daniel: So hereditability estimates are, I think, quite complicated to interpret. And, you know they are bounded by the particular context in which they’re observed and the quality of measurements that are made. And so, one of the reasons that estimates of the hereditability of human lifespan are very low is that there are many deaths caused by processes that are not aging and so that’s going to limit our ability to quantify how much genetic variance there is in human longevity. But in some ways I think that whether it’s 10 percent, or 20 percent, or 30 percent we can agree it’s tremendously important. and it’s also far from the whole story.

Gordon: Yeah.

Daniel: so to turn to your next question of [laughs] —

Gordon: Yeah.

Daniel: — how do we sort out the remainder of that variance whether it’s 90 percent, or 70 percent, or whatever it is.

Gordon: Right.

Daniel: I think there are a number of different approaches. Certainly, in the field of exposomics the goal is to survey variation in the environment in the same way we are now surveying variation in the genome by doing comprehensive assays of the composition of air, water, building materials. Incorporating as many measurements as we can into a full profile of the spaces and places in which people live… That’s a huge data challenge from the standpoint of assembling all the measurements. It’s a huge analytic challenge from the standpoint of sifting through those many correlated exposures that people encounter. But progress is being made and in the same ways that we’re making it as we try to sort out which of the several million genetic variants that we study has consequence for disease or the hundreds of thousands of methylation marks. In parallel I think that there is a very mature science in the social behavioral sciences, and it was recently recognized with a Nobel Prize a few years back, of quasi-experimental designs. Of research designs where we identify events in history, changes in policy, weather events or disasters that are beyond the control of the individuals that who experienced them. The Dutch famine that we talked about at the very beginning of the interview is an example of one of these things. And these quasi-experimental designs that compare individuals who were and were not exposed to these events provide a powerful method for determining the causality of a particular kind of exposure intersecting human lives at a particular point in development. And so, I think that as we move forward in identifying the ways in which environments shape aging, those quasi-experimental designs will be tremendously important. In parallel, we’re beginning to see the development of another class of studies that allow us to do the same kind of thing. And these are field trials where people go out into the world and change the way people live —

Gordon: Hmm.

Daniel: — or change the places in which they live.

Gordon: Mm-hmm.

Daniel: And a very simple design, um for something like this is the the unconditional cash transfer intervention which we’ve seen a lot of in development economics where-where people are simply handing cash —

Gordon: Yeah.

Daniel: — to people in low-resource conditions. but there are other kinds of interventions where we build water treatment facilities or sanitation facilities or otherwise attempt to improve the-the overall environmental conditions in-in which people live. Installing air purifiers. Or at the kind of macro level changing regulations around pollution to improve air quality at the level of a city or a region. And so, I think that those kinds of large-scale intervention trials will provide an important complementary resource to the natural experiment or the quasi-experimental design in figuring out those aspects of our exposome which we can manipulate to produce more healthy aging.

Gordon: You know, looking at the challenges it can be a little bit overwhelming. It’s kind of like climate change. You-you can accumulate the evidence, but it’s still an uphill battle to-to-to do something about it. But also, I sense optimism here in the way the different science expertise is coming together and focusing on aging itself. And you’re also talking about interventions that maybe we could be getting soon. What’s your take away on all of this right now?

Daniel: I think that we have within our power to change the environments we live in, in ways that can make profound differences in our health span. And I think we’ve already changed environments in ways that have made profound differences in our health span. We see it around us every day. It used to be the case that if you lived to be 100 the mayor of your town would give you a call, and you might even hear from your governor. And now, we have so many centenarians that they just can’t do it anymore- you’re going to have to make it to 105 or 110 if you want to talk to an elected official! That’s a reflection of the ways in which our public health apparatus, our medical apparatus have better supported health from the very beginnings of life all the way to its end. And the challenge we face now is how to expand access to all those extra years of healthy life that the most fortunate among us are able to enjoy. And I think that hopefully, with tools like the ones we’re building we’ll be able to rapidly identify the kinds of interventions that have the largest impact with evidence that is clear enough to sway policy in ways that can really help us build that better world.

Gordon: Wonderful. I’ll let you choose to be optimistic. I think we’re in a great place, and I think we’re going to see amazing things over the next few years. Daniel, thank you so much for taking time to talk to us today. This has been an eye-opener for me and I think I’m off to find out if I can get a PhD in epidemiology somewhere!

Daniel: Thanks very much, Gordon. It was my pleasure.

Gordon: Thank you.

Thank you for listening. Please subscribe, share and give us a five star review on Apple, Spotify or wherever you get your podcasts. We’re not getting any younger yet. Is produced by Vital Mind Media. The Buck Institute’s very own Robin Snyder is the executive producer, Wellington Bowler is right next to me here directing the recordings. Stella B is behind the scenes, ready to debrief when we up on the esteemed Sharif Ezzat weaves the show together for you. If you’re listening to the podcast, you’ll know that there has never been a more exciting time in the research on aging. discoveries in the lab, or moving to the clinic to help us all live better, longer. The Buck Institute depends on the support of people like you. To carry on our breakthrough research. Please visit us at Buck institute.org to learn more and to donate.

More episodes like this

Scientific Wellness and AI at the Forefront
What do we really want from our healthcare system, and how can AI help us get it sooner? In our final episode of season two, Gordon talks with visionary systems biology expert Nathan Price about the emergence of scientific wellness, what we can learn from our digital twins, and how using AI to predict health outcomes can help us transform our understanding of aging and disease—potentially adding years of healthy life...
Resilience as Medicine
Few people have as much experience with patient care and long-term research as the celebrated geriatrician and epidemiologist Luigi Ferrucci...
From Potential to Practice
The movement to integrate longevity treatments into clinical care is making strides...

Support the Buck

We rely on donations to support the science that we believe will add years to people's lifespan and decades to their healthspan.