Imagine you could take fourteen different ways of measuring how fast a person is aging — grip strength, blood pressure, inflammation markers, balance, memory tests, a DNA readout — and run them all in a fair race. The finish line is the bluntest one there is: which measure best predicts who dies over the following years? That experiment has now been done, and the winner surprised even the people who study these tests for a living. It wasn't a muscle test or a blood-pressure cuff. It was an epigenetic clock: a chemical fingerprint read from a single tube of blood.
The study that ran the race
The work came out in 2026 from the Berlin Aging Study II, one of the more carefully tracked groups of older adults in the world. Researchers took 1,083 people, aged 60 to 80 at the start, and followed them for an average of about seven and a half years.[1] They measured fourteen "biomarkers of aging" that an expert panel had recently agreed were worth testing in longevity trials — a mix of the physical (muscle mass, hand-grip strength, gait speed, a get-up-and-walk test, standing balance), the metabolic and inflammatory (blood pressure, C-reactive protein, interleukin-6, IGF-1), the cognitive, and the epigenetic.
Then they asked a hard question: adjusting for age, sex, lifestyle and genetic background, which of these actually predicted death? Most did not. Blood pressure, gait speed, muscle mass, IGF-1, C-reactive protein and several others showed no significant link to mortality in the adjusted models. Five held up — hand-grip strength, interleukin-6, standing balance, cognitive health, and an epigenetic clock called DunedinPACE. And of those five, the epigenetic pace-of-aging clock was the strongest and most consistent predictor, holding across different causes of death.[1] When the team hunted for the smallest set of tests that predicted death almost as well as all fourteen combined, they landed on just three: muscle mass, standing balance, and that same epigenetic clock.
What an epigenetic clock actually reads
To understand why a DNA test beat a grip dynamometer, you need to know what it is measuring — and, just as importantly, what it is not. An epigenetic clock does not read your genes themselves. Your DNA sequence barely changes across your life. What changes are the chemical tags sitting on top of it, the most studied being DNA methylation: small methyl groups (chemical formula –CH3) that clip onto specific spots on the genome and help decide which genes are switched on or off.
As you age, the pattern of these tags drifts in remarkably predictable ways. Back in 2013, the geneticist Steve Horvath showed that by reading methylation at a few hundred carefully chosen sites, an algorithm could estimate a person's chronological age to within a few years — across almost any tissue in the body. That was the first true epigenetic clock. Since then the field has raced ahead, and the most useful clocks stopped trying to guess your birthday at all.
Age clocks versus speed clocks
This is the distinction that trips most people up. The earliest clocks were age estimators: feed in a blood sample, get back "your body looks 58." Later "second-generation" clocks such as GrimAge were trained not on age but on health outcomes, and they turned out to be far better at predicting who would get sick or die.[3] GrimAge, built by Horvath's group in 2019, strongly predicted time-to-death, heart disease and cancer in tens of thousands of people.
DunedinPACE — the clock that won the Berlin race — goes a step further. Instead of a single snapshot age, it estimates your pace of aging: how many biological years you are burning through per calendar year. A result of 1.0 means you are aging at the expected one year per year. A 0.8 means you are running slow; a 1.2 means you are aging 20 percent faster than the calendar. It was built from the Dunedin study in New Zealand, where researchers tracked 19 different organ-system measures in the same people across two decades, then distilled all that change into a single blood test.[2] Think of the older clocks as an odometer and DunedinPACE as a speedometer — and for predicting what happens next, the speedometer often tells you more.
Why the number is more than a curiosity
A test is only as good as what it predicts, and here the epigenetic clocks have built a real track record. A large systematic review pulling together more than 150 studies found that faster epigenetic clocks were repeatedly linked to higher rates of death, cardiovascular disease, cancer and diabetes, and that factors like a high body-mass index, smoking and chronic stress tended to speed the clocks up.[6] The clocks don't all agree on the exact size of these effects — but they mostly agree on the direction, which is reassuring.
The pace-of-aging clocks add something the odometer versions can't: they seem to sit on the causal path between how you live and what happens to you. In the Framingham Heart Study, older adults who ate a brain-healthy Mediterranean-style diet had a slower DunedinPACE, and that slower pace statistically explained a meaningful share of the diet's link to lower dementia and death risk — roughly a quarter of the diet–dementia connection and more than half of the diet–mortality connection.[5] In other words, the clock wasn't just a passive readout; it behaved like part of the mechanism.
Can you actually move it?
This is the question everyone really wants answered, and here the evidence is genuinely promising but still thin. The strongest data come from CALERIE, a randomized controlled trial — the gold standard — in which 220 adults without obesity were assigned to either cut their calories by about 25 percent or eat normally for two years. When researchers went back and read the participants' epigenetic clocks, the calorie-restriction group showed a slower DunedinPACE pace of aging. Notably, the odometer-style clocks (PhenoAge and GrimAge) didn't budge — only the speedometer did.[4]
Two honest caveats belong right next to that result. First, the effect was small. Second, slowing a biomarker is not the same as adding healthy years to a life; no trial has yet followed people long enough to prove that moving a clock actually delays disease or death. The authors were careful to frame it as support for the idea that aging is modifiable — a real and encouraging finding — not as proof that dieting buys you extra birthdays. Observational studies point the same way for regular exercise and not smoking, but "linked to a slower clock" is a weaker claim than "shown to extend life."
The fine print the headlines skip
For all their promise, epigenetic clocks are not ready to grade your personal health, and the researchers who build them are the first to say so. Different clocks can disagree by several years on the exact same blood sample. Results wobble depending on the lab, the chip used to read the methylation, and even which version of the algorithm is run. A clock is impressively steady when averaged across a thousand people in a study — which is why it shone in the Berlin data — but far noisier when you try to pin a single number on one individual or track a small personal change over a few months.
That matters because a growing industry now sells direct-to-consumer "biological age" tests promising to reveal your true age from a finger-prick or a cheek swab. The underlying science is real, but none of these tests is cleared by regulators to diagnose anything or guide treatment, and a reading that swings by a couple of years between kits can send someone chasing an expensive supplement stack for noise. Treat any at-home result as a rough, motivational snapshot — interesting, not diagnostic.
An honest verdict
The epigenetic clock earned its moment. In a fair fight against thirteen other respected measures of aging, a pace-of-aging clock was the best single predictor of who lived and who didn't — a genuinely striking result that tells us these DNA marks capture something deep about the aging process. And unlike a lot of longevity hype, there's early randomized-trial evidence that the clock can actually be slowed.
But the sober reading is that epigenetic clocks are, for now, powerful research tools rather than personal health verdicts. They're brilliant for spotting which populations, and which interventions, age faster or slower. They're not yet good enough to tell you, specifically, how many years you have in the tank — and no clock has been proven to buy extra healthy time when you move it. The most reliable levers on your own pace of aging remain the boring, well-evidenced ones: don't smoke, stay physically active, eat mostly plants, sleep enough, keep a healthy weight. The clock is watching those choices closely. It just can't yet tell you your fortune.
Common questions
What is an epigenetic clock?
An epigenetic clock is a test that estimates biological age from chemical tags, called DNA methylation, sitting on your genes. These tags change in predictable patterns as you age, so an algorithm can read a blood or saliva sample and output either an age estimate or, in newer clocks like DunedinPACE, a speed — how many biological years you are aging per calendar year. It is a research tool, not yet a validated clinical test for individuals.
Can you slow down your epigenetic clock?
There is early randomized-trial evidence that you can nudge it. In the CALERIE trial, two years of moderate calorie restriction slowed the DunedinPACE pace-of-aging clock, though the effect was small and other clocks did not move. Observational work links a healthy diet, regular exercise and not smoking to slower clocks. No intervention has yet been shown to add healthy years by moving a clock, so treat it as a promising signal, not a proven lever.
Are direct-to-consumer epigenetic age tests accurate?
They are reproducible enough for research on groups of people, but far less reliable for telling one person their true biological age or tracking small personal changes. Different clocks can disagree by years on the same sample, results shift with the lab and the algorithm version, and none is FDA-cleared to diagnose or guide treatment. They are best read as a rough, motivational snapshot rather than a medical verdict.