Generic selectors
Exact matches only
Search in title
Search in content
Post Type Selectors
Generic selectors
Exact matches only
Search in title
Search in content
Post Type Selectors

A Deeper Dive into DunedinPACE

Listen or watch on your favorite platforms

Everything epigenetic
Everything epigenetic
A Deeper Dive into DunedinPACE
Loading
/

According to Dr. Daniel Belsky at Columbia University, there are three limitations of epigenetic biological age clocks:

  1. Mortality selection 

Essentially, biological age measures may underestimate true aging because older participants represent slower agers. 

  1. Cohort Effects

Biological age measures may overestimate true aging because older participants carry an excess burden of early-life exposure to environmental toxicants, pathogens, poor nutrition, smoking, etc. 

  1. Uncertain Timing 

Biological age measures summarize total aging over the lifespan and cannot distinguish differences established early in development from ongoing processes of aging. As a result, biological clocks may have lower sensitivity to effects of intervention. 

So, you’re probably wondering, how do we account for these limitations?

Dr. Belsky and his team have created a tool that enhances the precision of measuring the rate of biological aging. Their work involved observing the health outcomes of 954 participants across four different age groups spanning from the mid-20s to the mid-40s. The researchers examined biomarkers believed to indicate how well various organs are functioning, as well as others linked to general health. Using this data, they devised an epigenetic “speedometer” to forecast how these values would change over time. 

This tool is called the DunedinPACE.

As you may already know, the DunedinPACE measures how fast you are aging biologically for every one chronological year. If you need an introduction to DunedinPACE, check out my episode with Dr. Terrie Moffitt HERE.

In this week’s Everything Epigenetics podcast, Dan Belsky and I take a deeper dive into why Biological Age is limited and how DunedinPACE overcomes these limitations. Dr. Belsky speaks with me about a geroscience model of aging-related burden of disease, DunedinPACE test-retest reliability, and why the DunedinPACE indicates a faster pace of aging in individuals with older chronological and biological age.

We also discuss the effect of long-term caloric restriction on DNA methylation measures of biological aging in healthy adults from the CALERIE trial. 

The DunedinPACE is  a new tool for geoscience to investigate etiology in epidemiological studies and to evaluate the treatment effects of randomized controlled trials. 

Dr. Belsky continues to validate the DunedinPACE in other populations around the world.

In this podcast you’ll learn about:

– Dan Belsky’s unusual journey into aging science
– How to measure aging in younger people
– A geroscience model of aging-related burden of disease
– Why it’s important to have such model
– Clinical trials which increase healthspan in animal models
– Limitations of current biological age clocks
– Mortality selection/survival bias
– Cohort effects
– Variation in biological age clocks
– The retention rate of the Dunedin study
– The fifth round measurement of the Dunedin cohort
– The range of DunedinPACE (0.6 – 1.4)
– Why we see the DunedinPACE accelerated at older chronological ages
– The CALERIE RCT
– The value of DunedinPACE
– Dr. Belsky’s focus on public health approaches to promote healthy longevity

 

Dr. Belsky’s research sits at the intersection of public health, population & behavioral sciences, and genomics. 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.

Dan’s research in genetic epidemiology includes polygenic score studies of the development of obesity, asthma, smoking behavior, depression, and socioeconomic risk. His work in aging has focused on the development and analysis of algorithms to quantify the process of biological aging, especially in young and midlife adults. 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, the JAMA journals, Lancet Respiratory Medicine, and top journals in epidemiology and gerontology.

Dan is currently pursuing three related streams of research: (i) Development of methods to quantify processes of biological aging in young and midlife humans; (ii) Analysis of longitudinal cohort study and randomized trial data to identify molecular and behavioral pathways to resilience through which at-risk individuals can slow their pace of aging; and (iii) Analysis of gene-environment interplay to identify environmental factors that can be modified to reduce genetic risk for age-related disease and functional decline.


Belsky recently developed a novel DNA methylation measure test to quantify the pace of biological aging from a single-time-point blood test in collaboration with the Moffitt-Caspit Lab at Duke University.

About this guest 

Twitter

Dan Belsky’s profile at Columbia University: https://www.publichealth.columbia.edu/profile/daniel-belsky-phd

The Belsky Lab: https://www.belskylab.com/about

DunedinPACE study: https://elifesciences.org/articles/73420

CALERIE RCT: https://www.nature.com/articles/s43587-022-00357-y

Transcript:

hannah_went (00:01.031)In today’s episode, we talk with Dr. Daniel Belsky. Welcome to the Everything Epigenetics podcast, Dr. Belsky. We’re excited to have you.dan_belsky (00:09.078)

Well, thank you. Happy to be here.

hannah_went (00:11.619)

Yeah, definitely. So I know you’re involved with a plethora of research. When I started creating my podcast agenda to send over to, it was really hard to decide what I wanted to focus on. I’m definitely going to have to have you back on and, you know, a couple, a couple more years. But I’d love to hear a little bit more about you and your journey and how you became involved in this space. So can you give our listeners some information about that?

dan_belsky (00:40.014)

Sure, I mean, I think I have a somewhat unusual journey into aging science and the kind of work that brings me to you today. So I started out my career with a PhD in health policy. I was interested in ways to reform the healthcare system to address social and racial and ethnic inequalities in health in the United States. And while I was there, I discovered that on the one hand,

hannah_went (00:43.89)

I’m sorry.

dan_belsky (01:09.89)

Many of the inequalities I was interested in had their roots far earlier in human development Than people’s encounters with a health care system But also that I really didn’t have much to contribute in terms of the ways we could make that system better. So my intuitions Had more comparative advantage in thinking about what made people sick than how our health care system could be better organized to make them better And that led me to begin pursuing training in epidemiology first

hannah_went (01:27.499)

I’m sorry.

dan_belsky (01:38.278)

through research in human development and later through research in aging. So I had a pre-doctoral fellowship from a human development institute and I was studying children. And as I went on, I got interested in what was happening to those kids as they grew up and grew older, and that led me to a post-doctoral fellowship at the Duke University Aging Center where I learned about

biological aging as a modifiable cause of many different chronic diseases. And that really appealed to me as a new topic of inquiry because it matched the pattern of illness that I was seeking to understand and ultimately try and address, which is that kids who grow up poor get sick and die younger than kids who grow up rich. And that’s caused by essentially every different disease you can imagine. And here was a theoretical model proposing

a biological process of aging that underpinned all of those different diseases. And so I got very interested in how we might study this process and seek to modify it. But what I discovered early on was that there was a big problem in translating these animal models of aging as a cause of disease into humans, which was that we didn’t have any way of measuring aging in people. We could study older adults and we could observe their rate of onset of different chronic diseases.

conditions, their rate of deterioration in certain physical functions, and of course differences between people and how long they lived. But that was all work that was focused on humans who had already had exposures that accelerated their aging and after it was already in some ways too late to prevent that. And so if I wanted to move backward in developmental time to get there in time to help people live longer, healthier lives, we were going to need to understand how to measure aging in younger people.

hannah_went (03:13.657)

Mm-hmm.

dan_belsky (03:27.97)

And so with my mentors, Terry Moffitt and Avshalam Caspi, we set about developing measurements of biological aging that could be implemented in young people. And along the way, I met Morgan Levine and Alan Cohen and Steve Horvath, and I learned a lot from them about how to think about this problem and how to pursue this enterprise. And that’s sort of the chain of events that led me to where we are today.

hannah_went (03:53.383)

Yeah, super interesting. Thank you for that. You have a unique entryway, I would say, into the space and into the field. I think a lot of our listeners would be interested and we’ll get into the weeds of this, but about the…

the younger people and their aging, right? And looking at cohorts specifically within that. So I first want to address though, the subject of geroscience, right? You talk about that a lot in some lectures you’ve done and some other different webinars. So, and you say a geroscience model of aging related burden of disease. And I don’t know that that’s a phrase people are very familiar with. And that’s what we’re gonna start our discussion with today. So can you explain this to our listeners? Why is it important?

to have this type of model.

dan_belsky (04:39.95)

Sure, so the idea of geroscience is that biological changes that accumulate within and between ourselves as we grow older are modifiable causes of a range of aging related diseases. So epigenetic alterations, mitochondrial dysfunction, metabolic changes, dysregulated nutrient sensing, misfolded proteins accumulating within and between ourselves, all of these things.

hannah_went (04:55.262)

Mm-hmm.

dan_belsky (05:06.402)

progressively undermine the integrity and resilience capacity of cells, tissues, and organ systems, in turn driving vulnerability to all the different diseases we tend to get as we grow older. The Geroscience hypothesis emerged from experiments with worms, flies, mice, and sometimes other organisms, yeast cells, for example, showing that slowing or reversing the accumulation of these molecular changes could increase

hannah_went (05:27.104)

Mm-hmm.

dan_belsky (05:36.126)

lifespans of animals and in fact more recently increased health spans of animals. So not only do they die less often and at longer time removed, but in fact they maintain the functions that these animals perform in laboratory settings for longer periods of those lifespans. And that motivated interest in aging biology in the

potential to translate these discoveries to therapies that could slow human aging and extend healthy lifespan. And so that’s the, the geroscience model is basically the idea that aging begins with these cellular level alterations, feeds forward to a decline in the integrity of our organ systems and ultimately causes disease, disability and death. And then critically by intervening at that molecular stage, we can forestall or at least delay.

the decline in organ system integrity and preserve healthy lives.

hannah_went (06:35.195)

Sure, sure. And I wanna pick something out that you said there. You mentioned that more recently there was a study or something that came about that actually extended the health span, maybe in animal models. Can you talk a little bit about that? I’m not sure I’m familiar.

dan_belsky (06:49.302)

Yeah, so I think there are a handful of examples, but administration of the drug rapamycin, for example, keeps mice running on treadmills for longer than the mice who don’t get the rapamycin. You can see the same thing in caloric restriction experiments. So rapamycin is a drug. It’s called rapamycin because it was derived from a bacterium discovered in the soil of rapanuia, which is a

hannah_went (06:57.233)

Oh, gotcha. Okay.

dan_belsky (07:16.67)

Easter Island where the moai are. And we use it in human medicine to prevent rejection of organ transplants. So it suppresses the immune system in that way but it turns out to have this range of other properties as well. So it’s a drug and at the other end of the spectrum we have caloric restriction which is a behavioral intervention and we know when we give that to mice they also maintain functional capacity for longer periods of time and in fact this is also seen

hannah_went (07:18.701)

Mm-hmm.

dan_belsky (07:44.358)

in non-human primates, in the rhesus monkeys, in the National Institute on Aging, a long-term study of caloric restriction in which they show that the animals in the CR condition maintain muscle tone and integrity as well as organ system health much longer than the animals who are eating a normal diet.

hannah_went (08:03.355)

Understood. Yes, I think I misinterpreted that. So yes, I’m definitely familiar with those studies. And, you know, Dr. Matt Caberline is doing a lot of interesting research with the rapamycin and, you know, dog models. So we’ll keep understanding, I think, how exciting that is as it relates to, you know, hopefully switching some of those methylation markers in our favor, thus seeing the reverse of biological aging and some extension of health span. So super interesting. You know,

One of the problems though with healthspan that we were just discussing and following from the geroprotective interventions, that really takes too long, right? Patients, humans in general, are very impatient. So, you know, the gold standard for measuring these changes, I would say, used to be the biological age clock. I guess it really depends on the outcome or output that is of interest or what you want to look at. But can you describe some of the limitations with

biological clocks and then we’ll get into why the Dunedin pace is so exciting and it’s maybe ability to capture change a little bit better.

dan_belsky (09:07.662)

Sure. So again, the idea is that when we think about translating these

therapies from animals to humans, we have to also change the way we’re measuring outcomes because in worms, flies, and mice, we can run studies for months and watch the animals age, begin to develop aging-related disability and disease. But in humans, the same kind of studies would take decades. And for that reason, people have proposed using measurements of these biological processes of aging or surrogates for them as what the Food and Drug Administration here in the United States calls surrogate endpoints of…

hannah_went (09:15.956)

Mm-hmm.

dan_belsky (09:42.542)

treatment effectiveness. So the idea is that instead of having to follow the individual to determine healthy lifespan, we could take these biomarker measures and they would give us a near-term readout of the interventions impact on that future healthy lifespan. And there have been a number of these proposed. And I think that most of the modern approaches to developing these surrogate endpoints have focused on integrating large numbers of molecular measurements using

machine learning algorithms. And many of them have used those machine learning algorithms to design measurements that produce values that are comparable to chronological ages. And so we call these biological clocks and we call the measurements that they produce biological ages. And the idea is that the difference between the biological age estimated by the biomarker algorithm and your chronological age is a measurement of how

hannah_went (10:34.293)

Thanks for watching!

dan_belsky (10:38.574)

much faster or slower your aging or how much more or less you’ve aged than we would expect given your chronological age. And there are a couple of different ways that people develop these biological clocks. The most straightforward and intuitive of them is they simply compare older and younger people and use the machine learning algorithm to parameterize differences in the biology between the older and younger people to predict their age.

hannah_went (10:55.595)

Mm-hmm.

dan_belsky (11:04.014)

And so those age prediction models formed the first epigenetic clocks, and there are clocks from other molecular substrates that have been used in that same way. And in the case of DNA methylation, these age prediction clocks have proven to be astonishingly accurate. So Steve Horvath’s 2013 clock is probably the most famous of these. It ticks in almost every cell in the body.

and can do a pretty impressive job of estimating how old you are, including from your blood samples, which are readily available to researchers in clinical trials, and which is not a small accomplishment given that your blood cells have probably only been around for a few weeks, and yet somehow these chemical tags on the DNA sequence of their nucleus are able to recover how long it’s been since you were born. So there’s clearly some fascinating biology bound up in these age,

prediction clocks. But it’s turned out that they’re not that predictive of differences between people in health, in future onset of disease or disability or lifespan. And it also turns out that they’re not that well correlated with patterns of physiological differences between people that we think are that intermediate stage in the geroscience model. So from cellular level changes to organ function to disease and disability. And that motivated

the clockmakers to innovate new approaches. The first example is really Morgan Levine’s phenotypic age clock that was published back in 2018, so five years after Steve’s initial paper. And in that approach, instead of looking sort of backwards to how long it’s been since you’re born, the clockmakers are trying to look forward in time to how long it’s going to be until you die. And so in the design of these studies, what you’re doing is you’re measuring people at some baseline.

often at different chronological ages, and then you’re following them for a fixed period of time and comparing differences in survival. And the machine learning algorithm is then parameterizing those survival differences to generate a prediction of a mortality risk. And that mortality risk is then converted into an age value based on what the typical mortality risk is for each chronological age in some reference population. So that innovation dramatically improved

dan_belsky (13:25.026)

the capacity of these epigenetic clocks to predict disease disability and mortality. And one of the reasons we think that that’s the case is it addressed a pretty important confound in the age prediction clocks, which is that when you compare younger and older people, you’re comparing what are effectively two different populations. Young people represent the entire population with no attrition from age-related morbidity and mortality.

hannah_went (13:54.293)

Yep.

dan_belsky (13:55.006)

more or less everybody. So we don’t have zero infant mortality in this country. We don’t have zero childhood and young adult mortality, but mortality due to aging related diseases is very close to zero at those age ranges. In contrast, if we move forward to people in their 60s, 70s and 80s, those aging related disease has become a primary cause of attrition from the original birth cohort.

hannah_went (13:56.843)

Thanks for watching!

dan_belsky (14:18.846)

And so what our machine learning analyses are doing essentially then is comparing the average young person to successful agers at the older end of the lifespan. And what that’s going to do is attenuate the biological information about aging we’re able to recover from those algorithms. Interestingly, when people use the same approach to develop epigenetic clocks in mice, what they find is that those epigenetic clocks are much more effective at predicting patterns of morbidity, mortality or healthy lifespan. And it may be because

in the mouse models, they’re able to control environmental conditions and cohort attrition so that they know the differences in age are in fact going to pattern with biological risk for disability, disease, and mortality. In any case, so mortality selection or survival bias is the first of the limitations of the original epigenetic clocks, and that’s largely addressed in the second generation of epigenetic clocks that predicts mortality.

hannah_went (14:52.348)

Yeah.

dan_belsky (15:15.914)

potential limitation of epigenetic clocks are what demographers call cohort effects. And these are differences in exposure histories for people born at different points in history. And so the mortality prediction clocks solve the problem of survival bias, but they still end up comparing older people and younger people, and they still end up assuming that the differences in risk of death are attributable to differences in biological processes of aging. However, if there are causes of disease and death…

hannah_went (15:21.382)

Mm-hmm.

dan_belsky (15:45.942)

that are not mechanisms of aging, but that do leave their imprint on, in this case, the methylome, they will be incorporated into these algorithms as part of the biological clock. And while that may be fine for the purposes of predicting who’s going to die sooner, it may be a problem if we wanna use these measurements as surrogate endpoints in clinical trials where our interventions modify biological processes of aging, but perhaps don’t reverse damage induced by chemical toxic and exposures.

hannah_went (15:58.449)

Mm-hmm.

dan_belsky (16:15.954)

or pathogens early in life. And that brings us then to the final potential limitation of the biological clocks that at least that we’ve focused on, which is that we don’t know when the variation in biological age that they measure came about. So you may take a DNA methylation measurement of a person, it reads out a biological age, that biological age is older than the person’s chronological age, suggesting some accelerated aging process. But it’s not clear whether that…

advance in biological age was generated gradually over time, consistent with accelerated aging, or instead reflects some insult or injury earlier in the lifespan, which is unlikely to be ameliorated by some anti-aging therapy. And so that’s what brings us to the approach that we took in developing our measurements at the pace of aging and ultimately Dunedin pace. And so the design of this method was built around addressing those three limitations of

hannah_went (17:08.629)

Yeah.

dan_belsky (17:14.014)

survival bias, cohort effects, and uncertain timing in determination of variation.

hannah_went (17:20.263)

Yeah, absolutely. I think that, again, to summarize, there’s really those kind of three limitations you mentioned, the survival bias, which the second generation clocks really accounted for then. But still in question was the cohort effects and the variation in the biological age measurement, when they have acceleration or the…

deceleration of that aging occur in that lifetime. And I think number three is probably most frustrating for people who are taking a lot of these commercialized tests that are available, right? Because if they see accelerated aging, everyone thinks of biological age. They think of this age and number, and they say, oh no, and make it upset or freak out and think that happened right then. But it could be something that has happened possibly in the past. And that’s why I love the Dunningdon pace so much is because it actually tells you

currently aging. It gives this further insight into the aging process, but you know again I think when people think biological age they want a number they are more familiar with that and there’s some education with the Dunedin pace. So

I always tell people, you know, when I talk about this with a lot of healthcare providers and discuss these types of clocks, the Dunedin cohort is the very first of its kind and it’s such a unique cohort. I definitely hope it’s not the last. Is there any other type of study like that out there or any other cohort?

dan_belsky (18:45.866)

Well, so the Dunedin longitudinal study is a birth cohort study. So it enrolled babies born in a particular hospital in Dunedin, New Zealand between 1972 and 1973, and it has followed them up over the subsequent nearly 50 years. To the question of whether there are other studies with a similar design, there are many birth cohort studies out there in the world. Few of them have measured their subjects as intensively.

hannah_went (18:51.697)

right.

hannah_went (18:55.353)

Mm-hmm.

hannah_went (19:02.421)

Mm-hmm.

hannah_went (19:09.276)

Right.

dan_belsky (19:14.75)

as the Dunedin study. And what really makes the Dunedin study unique is its success in retaining the original participants over this very long duration of follow-up. So in a typical cohort study, there’s very substantial attrition early on and then continued attrition as you move forward in time. So you might see a 60% retention rate at your second follow-up, and then only 80 or so percent of those are retained at each subsequent.

follow up from there. The Dunedin study, across its nearly 50-year lifespan, has succeeded in maintaining the participation of more than 90% of the surviving members of the original birth cohort. And that was achieved in part through the efforts of the study to build a goodwill and trust with the mothers of the study participants, because when they were first enrolled, of course, they were babies, and it was their mothers who were choosing to participate.

hannah_went (19:57.722)

Mm-hmm.

dan_belsky (20:13.518)

But also through the tireless efforts of the study directors, now Richie Poulton in Dunedin, New Zealand, and then Terry Moffat and Ashlam Kasbi at Duke University, who’ve worked tremendously hard to ensure that participants who may not respond to phone calls or mailings are followed up again and again to the point where they will even send research workers to find people living in far-flung places so that they can conduct interviews and maintain participation in the study. And that was motivated in part

hannah_went (20:38.318)

Mm.

dan_belsky (20:43.162)

by an interest early on in the study in tracing the natural history of psychiatric disorder in the general population. And of course, people who develop mental disorders may be somewhat less likely to continue participating in research studies than others. But it’s become a central focus in the interest of ensuring that age-related disease and disability don’t cause declines in participation and therefore a loss of that critical range of variation in the population.

hannah_went (20:50.077)

Mm-hmm.

hannah_went (21:09.375)

Gotcha, understood, very interesting. Yeah, I’ve heard somewhere between like a 94, 96% retention rate, which is amazing. It’s insane. And yeah, I spoke with Dr. Moffitt not too long ago. Her and Dr. Caspi are on their way back to Dunedin, correct? To collect another round, that fifth round of data. Perfect, yeah, I’m excited to see. So you all with that fifth round of data will do kind of an update or upgrade to the Dunedin PACE algorithm. What will that look like?

dan_belsky (21:24.442)

They are in New Zealand right now. That’s right. Yeah.

dan_belsky (21:37.07)

Well, I think that’s the hope. And I think you would need to talk to Terry about what her plans are. But I think that part of the vision for the next phase of Dunedin PACE is to move beyond a single composite predictor of aging at the whole organism level and to begin exploring whether predictions of aging within specific organs or systems can be measured.

hannah_went (21:43.871)

Mm-hmm.

hannah_went (22:04.203)

Gotcha, gotcha, understood. Moving on to one of the most asked questions that I receive and I honestly don’t know the answer to, I was excited to ask you this one. The range you can get when you’re measuring your dening-den pace is between 0.6 and 1.4. Can someone get a lower pace of aging than 0.6 or higher than 1.4, why is it that range?

dan_belsky (22:29.438)

I don’t think there’s anything fixed about the range of values that are being observed. So I think maybe it’s worth taking a step back and thinking about what the pace of aging is to begin with. So when we originally developed pace of aging, the idea was to see if we could track how fast people were aging over months and years of follow-up. And so you might think of a biological clock or these biological age measurements as

hannah_went (22:34.929)

Mm-hmm.

Yeah.

Yeah, yeah.

dan_belsky (22:57.258)

your car’s odometer, they tell you how many miles you’ve traveled, and the diddian pace as the car’s speedometer, it tells you how fast you’re going. And so to produce that speedometer, what we did was measure indicators of the function or integrity of multiple different organ systems in the body at repeated occasions over in the first study 12 years and then subsequently over 20 years of follow-up. And for each of these individual measurements,

hannah_went (23:22.351)

Mm-hmm.

dan_belsky (23:24.79)

We compared values at first three and then four different time points to determine how fast change was occurring in each individual participant. Once we were able to model those biomarker specific rates, we composited the rates across the biomarkers to form a single index. And we normalized that index to have a value of one in the average man and a value of one in the average woman.

hannah_went (23:35.901)

Mm-hmm.

dan_belsky (23:51.402)

representing the amount of physiological change that we expect to occur in the passage of a single 12 month interval during the follow-up period. So that’s what the value of one means. It’s the expected biological change over 12 months of follow-up. And a value below one or above one then indicates less than or more than the expected amount of change. And we can think of that in percentage terms. So if your pace of aging is 1.1, it means you’re aging 10% faster.

hannah_went (24:00.849)

Okay.

dan_belsky (24:17.726)

If your pace of aging is 0.9, it means you’re aging 10% slower than the average participant in the study. And that’s how we interpret values of Dunedin pace. So we produce Dunedin pace by taking those same machine learning algorithms that are used to produce the epigenetic clocks and pointing them at pace of aging. And pace of aging is now completely differentiated from chronological age because everybody in the Dunedin birth cohort is exactly the same chronological age. They were born in the same calendar year.

and they are measured as close as they can get to their birthday. And so when we draw that blood sample, we’re drawing that blood sample from people who are all the same chronological age. And the only variation that is generated in pace of aging then is differences between people of the same chronological age and in how fast they’re aging. So the Dunedin pace values that range from 0.6 to, you think you said 1.4, represent variation in the aging rate in the Dunedin study.

hannah_went (24:47.299)

Mm-hmm.

hannah_went (24:59.714)

Mm-hmm.

hannah_went (25:14.626)

Mm-hmm.

hannah_went (25:17.855)

Mm-hmm.

dan_belsky (25:18.798)

from essentially 40% slower than the norm to 40% faster than the norm. But in all humans, it’s possible that values could exceed that range. You know, so here in the Dunedin study, they’re looking at just a thousand people. But in larger cohorts, we might expect more variation. So there’s no.

dan_belsky (25:42.994)

empirical upper or lower bound to Dunedin pace, although we might think that values that are a great deal slower, a great deal faster than that, you know, would be implausible or might suggest artifacts in the data rather than true variation in aging.

hannah_went (25:47.272)

Sure.

hannah_went (26:00.479)

Gotcha, understood. Thank you for that. I’ll be able to relay that information now to everyone who asks. And this may be a better question for Dr. Moffitt, but something that I just thought about as well that a lot of people ask me is, within the Dunedin cohort, are you telling that cohort to do anything or you’re not? They’re just living their life as normal, correct?

dan_belsky (26:24.438)

That’s right. So these are questions for Terry, but the Dunedin study is an observational cohort study. It is not an intervention. And so participants are measured. If in the exams of the patients, which now include brain scans, abnormalities are detected that are clinically significant. That information is shared back with the patients or with the participants because that’s the duty of the study.

hannah_went (26:27.078)

Okay.

hannah_went (26:30.812)

Yeah.

hannah_went (26:35.583)

Perfect.

dan_belsky (26:54.526)

But otherwise, the participants are not instructed to perform any particular change in behavior or lifestyle.

hannah_went (27:04.479)

Perfect. Just wanted to make sure that was clear as well. So, you hinted at this a little bit earlier in measuring the aging in younger people, but why may we see the Dunedin pace accelerated at older chronological age values in humans?

dan_belsky (27:22.03)

Sure, so one of the observations about Dunedin pace is that even though the measurement was devised in people who are all the same chronological age, it tends to be correlated with chronological age in studies of people who are of different ages. So older people tend to have a faster pace of aging than one, younger people tend to have a slower pace of aging than one. And our interpretation of this is that

hannah_went (27:31.776)

Mm-hmm.

dan_belsky (27:48.53)

it reflects a well-known principle in bio-demography, which is that the risk of death accelerates as we grow older. So what’s called the Gompertz law describes this pattern. From the age of 30 or so, your risk of death doubles every eight years about. And we see that same pattern of exponential increase in risk for a range of age-related conditions. Alzheimer’s disease, cardio…

chronic obstructive pulmonary disease, even common conditions like heart disease and diabetes showed this nonlinear pattern of increase with advancing age. And so our interpretation of the faster pace of aging that we observe in older people is that it is consistent with this hypothesis that the rate of aging itself is accelerating as we grow older. And it’s worth noting that simple damage accumulation models show the same pattern.

hannah_went (28:29.803)

Mm-hmm.

dan_belsky (28:47.298)

So if you look at the survival of non-biological systems, they also show that same exponential decline and an increase in failure rate. And so that’s what we think is being reflected in the accelerated values of Dunedin PACE. But I think it’s worth.

being clear that we’re in the early days of understanding what these measurements mean. And so it could be the case that some of the variation in Dunedin pace is reflecting differences in immune stance between people whose bodies show more rapid trajectories of deterioration versus slower trajectories of deterioration. And those differences in immune stance may increase as we grow older.

hannah_went (29:15.007)

Thank you.

dan_belsky (29:41.066)

what the Dunedin pace is picking up at older ages is simply a changing immune stance in the blood samples of individuals. And as we become more sophisticated in disentangling aging-related epigenetic change within cells from aging-related changes in the populations of cells in our blood, we’ll be able to answer that question.

hannah_went (29:41.739)

Mm-hmm.

hannah_went (30:08.667)

Absolutely. Yeah, I never had that insight. So interesting to think about. Another study you hinted at earlier, the calorie randomized control trial. So, you know, backing up and asking the question beforehand, is the Dunedin pace able to tell us whether an intervention is able to actually slow biological aging? So can you give us some insight into that and maybe the calorie randomized control trial as well?

dan_belsky (30:32.31)

Sure, so, you know, as we mentioned before, a key application of these measurements of biological aging is to serve as surrogate endpoints in randomized controlled trials of interventions that are designed to slow or reverse the accumulation of these hallmarks of aging and extend healthy lifespan. In model systems, one of the most effective interventions for accomplishing that

hannah_went (30:40.863)

Mm-hmm.

dan_belsky (31:00.846)

is caloric restriction and that’s a reduction in macronutrient intake with maintenance of nutrient sufficiency. So it’s not starvation. It’s something that people with the assistance of healthcare professionals can achieve and there are people who live on calorically restricted diets. The National Institute on Aging was interested in whether the kinds of effects that could be produced in model organisms through caloric restriction could be

hannah_went (31:05.036)

Mm-hmm.

dan_belsky (31:29.954)

generated in humans. And so they designed the calorie study to test that hypothesis. And there was a phase one trial that was designed to figure out how we can calorically restrict people. And then the phase two trial was a safety and efficacy trial that was designed to determine whether long-term caloric restriction could be safe for humans and how effective different dietary regimes would be in achieving this. Or whether…

hannah_went (31:33.14)

Mm-hmm.

hannah_went (31:43.819)

Thanks for watching!

dan_belsky (31:59.006)

modification of diet could achieve this kind of caloric restriction over the long term. And I discovered the study while I was at Duke in the aging center as a postdoc and later joined the biorepository for the study. So the trial concluded in 2009, but a biorepository was formed to store the data and the biomaterials collected from participants over the course of the trial. And we conducted a study back in 2017.

in which we use data that were accumulated as part of safety labs that were collected from patients, essentially blood tests that were given every year to determine whether the intervention was harming the participants, to compute measurements of biological age and to ask whether the intervention slowed the increase in biological age over the two years of the intervention. And the finding in that study was that in fact it did. So

hannah_went (32:54.335)

Mm-hmm.

dan_belsky (32:55.25)

With that evidence in hand, we went to the NIH and said, we’d like to get some of those blood samples out of the freezer and analyze the DNA to understand if this intervention, which showed signs of slowing aging at the level of physiology, that middle stage of the geroscience model, was having an effect on the cellular or molecular changes at the beginning of the geroscience model. And they agreed to support that research activity.

hannah_went (33:20.171)

I’m going to go ahead and turn it off.

dan_belsky (33:24.554)

And so we conducted this study, which I think there’s a preprint out on the MED archive and the paper will be forthcoming in the journal Nature Aging sometime this spring.

dan_belsky (33:37.506)

The design of this study was to compare measurements of Dunedin PACE and several epigenetic clocks at pre-intervention baseline and at the follow-up assessments conducted at 12 months and 24 months. So there’s no post-treatment data available from CalRIE. Cy Doss at Tufts University and her collaborators are leading a follow-up that will collect biospecimens from these participants a decade post-intervention to find out whether there were lasting effects.

hannah_went (34:05.067)

Cool.

dan_belsky (34:05.642)

but for the time being all we know is what happened during the trial. What we found in that analysis, which was led by some postdocs in my group, first Reem Waziri and now more recently Kaylin Ryan, is that the intervention did not change the rate of increase in any of the epigenetic clocks that we studied, but it did slow Dunedin pace.

from baseline to 12 months, and that slowed pace of aging was maintained through the second 12 months of follow-up. So the magnitude of the effect was small. It’s equivalent to, you know, we would interpret as a 2% reduction in the pace of aging. That’s not a trivial reduction. It corresponds to maybe a 10% reduction in the mortality hazard. And is…

hannah_went (34:36.139)

Mm-hmm.

hannah_went (34:41.602)

Thank you.

dan_belsky (35:02.338)

commensurate with the kinds of effects that people find in trials of smoking cessation. Nevertheless, it’s not as large an effect as we might’ve anticipated. And given how challenging caloric restriction is, it may not seem terribly satisfying to people who are thinking about trying it themselves.

hannah_went (35:16.034)

Mm-hmm.

I’m gonna go.

dan_belsky (35:24.578)

So the question then is why is it that we see effects on denedin PACE and not on, for example, the DNA methylation GrimAge, which is, if anything, more predictive of morbidity and mortality than PACE itself. And the answer is we don’t know. But what we speculate is that it has to do with the information content of these different measures. So GrimAge was designed to predict mortality differences between people.

hannah_went (35:41.567)

Mm-hmm.

dan_belsky (35:52.326)

from a cross section of biological data, from a single blood sample. And the information in that blood sample about risk of death reflects an accumulation of changes across the entire lifespan. The Grimmage was developed in the Framingham Heart Study, which measured DNA methylation in blood cells of their participants when they were, most of them were in their 50s, 60s and 70s. And so we have a half a century or more of living

that is bound up in that DNA methylation information and that is then being parameterized to predict mortality. And it could be the case that two years of caloric restriction just doesn’t put that much of a dent in a 50 year history of aging. And therefore the information that GrimAge was designed to capture is simply not as susceptible to intervention as for example, something like Dunedin PACE, which is designed to capture

on a year-to-year basis. But that’s a hypothesis, and it’s one that we have not yet tested. So at this point, it’s pure speculation. And it could be the case that it was random chance that we found an effect in one of these interventions and not the others. And I just want to emphasize that because the effects we detected in the trial were small, it’s quite possible that there are effects on something like grim age, but that in the case

hannah_went (36:52.735)

Sure. Yeah. Right.

hannah_went (37:05.593)

Mm-hmm.

dan_belsky (37:20.446)

a small trial of 200 people, they’re simply not detectable. Measurements that are more precise or trials that have a larger size will ultimately be needed to refine understanding of these kinds of small effects.

hannah_went (37:35.823)

No, I appreciate your thoughts in the speculation there. And that, just to confirm too, is that like about 10 to 11% overall caloric restriction somewhere within that range, I believe.

dan_belsky (37:44.206)

So the prescribed dose in the calorie trial was 25%, which is by the standards of mouse caloric restriction studies, not that severe, but by the standard of human living, really, really hard. And so what happened in the calorie trial was that

hannah_went (37:49.383)

Yep.

hannah_went (37:55.715)

Yeah. Yes.

dan_belsky (38:02.362)

Most people did not achieve the prescribed dose. In fact, by the end of the trial, almost no one did. On average, over the two years of follow-up, participants achieved 11 to 12% caloric restriction. And we did some analysis comparing people who were more successful versus less successful in reducing their caloric intake. And we did find that people who achieved a higher degree of caloric restriction showed a larger magnitude effect. So there’s an even pace.

hannah_went (38:09.061)

Right.

hannah_went (38:15.006)

Okay.

dan_belsky (38:31.198)

was reduced by a larger magnitude commensurate with that increased caloric restriction. We did some work to try and estimate the effect that would have been achieved in the case of a 20% reduced caloric intake, so roughly twice the average dose. And we do this using information from people who did get or who were that successful in reducing their caloric intake. And we find is that.

hannah_went (38:41.363)

Mm-hmm.

hannah_went (38:46.739)

haha

dan_belsky (38:59.722)

The effect is about twice the size of the one that we estimate for in the intent to treat analysis for the trial overall. So that would be, you know, on the order of like a 4% reduction in pace of aging corresponding to maybe a 20% reduction in risk of death.

hannah_went (39:02.398)

Okay.

hannah_went (39:12.171)

Sure.

hannah_went (39:18.559)

Perfect, yeah, has that made you restrict your calories? Does that give you any type of motivation? I know, like you said, it’s not the easiest intervention by any means. You know, eating is like this social habit and you know, people love to eat. They enjoy it, they indulge. So has that made you change your outlook at all?

dan_belsky (39:33.094)

Yeah, yeah. I, no, I mean, I don’t practice caloric restriction. And unlike many people in this business, I don’t undertake intermittent fasting either, which is now quite common among the geroscience crowd. But but I think that intervention studies like calorie provided some insight into the potential to

hannah_went (39:42.532)

Yeah.

hannah_went (39:48.596)

Right.

dan_belsky (40:01.51)

modify aging processes through behavioral modification. They can give us insight into some of the mechanisms and they can, I think most importantly in the context of our study, provide benchmarks for what is feasible in trials of therapies that may be easier for people to take up like intermittent fasting or regulation of circadian rhythms or even…

Pharmaceutical interventions, most immediately, the drug metformin is gonna be trialed as an intervention on aging. But looking to the future, there’ll no doubt be trials of senolytic therapies, of rapamycin therapy, and of other compounds.

hannah_went (40:31.68)

Sure.

hannah_went (40:42.324)

Yeah.

dan_belsky (40:44.338)

You know, so what Calory teaches us is that even a relatively extreme intervention like caloric restriction achieves relatively modest effects on these biomarkers. So either we need better biomarkers, bigger trials, or more effective interventions. And we’ll be working on all three of those things as a field, and we’ll see which one of them proves to be the most important. I think that it’s likely that what we see in Dunedin PACE is…

hannah_went (41:00.87)

Right.

dan_belsky (41:13.354)

maybe a lower bound estimate of the effects of something like caloric restriction. These DNA methylation measures of aging are far from perfect and far from being without measurement error. Although, we’ve designed PACE to have high technical reliability, meaning it’s pretty reproducible. If I test your blood again and again, I’ll get the same result. But that doesn’t mean it’s a perfect measurement of your pace of aging. There’s a lot that it’s not able to see. And so as we move forward,

hannah_went (41:17.465)

Mm-hmm.

hannah_went (41:31.004)

Mm-hmm.

dan_belsky (41:42.526)

and refine these measurements, we may develop the ability to better detect effects of interventions like caloric restriction on aging. And so part of the value of our study, in fact, was not the immediate test of does it modify gene even pace or grim age or one of these other epigenetic clocks, but in developing a database that will allow researchers to come back year after year as new measurements are developed and test the effects again and again. And so that will help clarify whether…

hannah_went (41:58.863)

Mm-hmm.

hannah_went (42:07.743)

Sure.

dan_belsky (42:09.058)

the relatively small effects we observe in calorie are down to our measurement approach or in fact the property of the intervention itself.

hannah_went (42:17.519)

Sure, yeah. So we’ll stay tuned for those studies to come. But I had to ask, I’m sure, people are thinking the same thing.

So switching gears, we’re almost getting toward the end of the podcast here. I know we talked about the limitations of those biological age clocks, those three points. So can you just explain or maybe summarize in a concise fashion, those who are interested in maybe testing their aging process, just maybe why you would favor that Dunnington pace over those biological age clocks, if there’s any other insight there just besides those three points.

dan_belsky (42:52.414)

Yeah, I mean, I think that anyone who’s doing this should be doing it primarily out of curiosity and not with an eye to tailoring their own behavioral regime. And I would say that if you are curious how your lifestyle or whatever you’re doing may be affecting your aging rate, then you should be measuring it in lots of different ways.

hannah_went (43:01.823)

Mm-hmm.

hannah_went (43:08.247)

Mm-hmm.

dan_belsky (43:21.358)

And so if you’re going to the trouble of paying to have DNA methylation analyzed in your tissue samples, you can very easily get all of these measurements back. I think pace of aging as a consumer product, to the extent that it is valuable, would be valuable in its ability to give a readout on how fast you’re aging now rather than how much you’ve aged in the past. But I would just have to encourage caution.

hannah_went (43:32.543)

Mm-hmm.

hannah_went (43:43.967)

Mm-hmm.

dan_belsky (43:49.642)

in interpretation on the part of consumers and consultation with healthcare professionals about any changes in lifestyle or other anti-aging interventions you might think about taking up before going forward. And in parallel, interpreting these measurements cautiously, because people do get epigenetic clock values that are implausibly high or implausibly low, and we don’t yet know what it means when…

hannah_went (44:03.156)

Right. Absolutely.

hannah_went (44:15.121)

Mm-hmm.

dan_belsky (44:17.898)

when we get that out, it’s very much the case that these algorithms are black boxes. We don’t know why the specific CPG sites that are used to compose the algorithms are selected other than that they collectively are predictive of the outcome. The need in pace is built on a unique outcome that tracks the rate of aging and that is something that is in theory a clinically valuable quantity.

hannah_went (44:23.922)

Mm-hmm.

hannah_went (44:35.956)

Mm-hmm.

dan_belsky (44:47.766)

But the precise interpretation of Dineen Pace values from the standpoint of personal wellbeing, I think is really something we don’t have a lot of knowledge about yet.

hannah_went (45:00.839)

Right, we’re still learning what these markers mean, right? I know there’s a lot of research behind the casual effects of the CPGs rather than just the correlation. So a lot of research being done in that space and something that we’ll look out for as it comes out. So Dr. Belsky, what’s next for you? What are you currently studying? What are you looking at? What are you excited about? I know that list is probably longer than we have time for, but yeah, what’s next?

dan_belsky (45:22.69)

Yeah.

dan_belsky (45:26.65)

So what are we doing next? So I think, again, unlike most of the people you talk to, our sites are set not so much on building next generation drugs to slow aging, and more on public health approaches to promote healthy longevity. And so what we’re doing now, in addition to continuing to develop and refine measurements of the aging process is,

to study how interventions that change the circumstances in which people live may affect their pace of aging. So the National Institute on Aging has supported us to go out and collect blood samples from people who participated in an economic intervention called My Goals, which promoted employment in people who were unemployed and living in public housing to find out whether…

hannah_went (46:02.964)

Mm-hmm.

hannah_went (46:13.681)

Mm-hmm.

hannah_went (46:20.494)

Wow.

dan_belsky (46:21.65)

improvements to people’s social and economic circumstances may modify their pace of aging. And it’s in some ways a long shot to see if this relatively brief but effective economic intervention has already altered the pace at which people are aging. But it will again provide us with a benchmark for the design of future trials and for our vision of what may be possible through policy to promote healthy longevity.

So that’s a big focus in our work and we’re going in that direction. And then the other thing that we’re working on is whether the pace of aging can be meaningfully measured and modified in young people. And so that’s a longer term project.

hannah_went (46:51.84)

Very cool. Yes.

hannah_went (47:06.983)

Awesome, well I’m excited for those to come. So the very last question I have for you, this is something I ask on the end of every single podcast, so this is kind of out of the blue. If you could be any animal in the world, what would you be and why?

dan_belsky (47:25.534)

If I could be any animal, what would I be and why? So I think, like without really giving it any serious thought or consideration, I immediately went to, you know, some kind of a bird because I think I’d like to know what it’s like to fly. But that’s without any serious consideration of what bird life is really like. I have to be honest when I think deeply about what it would be like to be some animal, none of it sounds that good.

hannah_went (47:30.348)

Yes.

hannah_went (47:36.479)

Yeah.

hannah_went (47:47.449)

Hahaha

dan_belsky (47:54.513)

I quite like being a human.

hannah_went (47:54.559)

Ha ha ha.

There you go. Yeah. Haven’t had anyone say bird yet, so that one’s unique, but it would be interesting to understand how to fly. Well, I really appreciate you and your time. So we’ve come to the end of this amazing podcast for people who may have other questions. I’ll link some show notes and a lot of those pre-prints and your papers as well. But in general, where can people find you if they have questions or wanna learn a little bit more about your research?

dan_belsky (48:22.582)

Yeah, sure. I’m at Columbia University’s Mailman School of Public Health. If you just Google my name, you’ll find my faculty webpage. And I think you can also find me at bellskylab.com. And I’m reachable by email, which you can find at either of those places.

hannah_went (48:34.88)

Perfect.

hannah_went (48:39.751)

Awesome. Thanks, Dr. Belsky. And to everyone listening, thank you for joining the Everything Epigenetic podcast. And remember you have control over your epigenetics. So tune in next time to learn how. Thanks again, Dr. Belsky.

dan_belsky (48:42.062)

Of course.

dan_belsky (48:52.627)

Okay, bye bye.

About this Guest Expert

Daniel Belsky
Dan Belsky, PhD, focuses on how genes and environments shape health across the life course, with an emphasis on developing and evaluating methods to quantify biological aging processes and exploring interventions to promote healthy aging and reduce aging-related social inequalities.

More About me

Everything epigenetic
Everything epigenetic
A Deeper Dive into DunedinPACE
Loading
/

More Episodes