The Epigenetic Clock: What Biological Age Measures

Overview

The idea that your birthdays understate or overstate how old you really are is not marketing. Epigenetic clocks are legitimate scientific instruments, they predict disease and mortality better than chance, and the underlying biology is well established.

The complications begin as soon as you ask for a number. There is no single epigenetic clock — there are dozens, built for different purposes, trained on different outcomes, and they do not always agree with each other when run on the same person's DNA.

The famous claim that lifestyle changes can reverse biological age by roughly three years traces to one pilot trial of 43 men over eight weeks. Its headline result was statistically significant against controls; its within-group result was not, and the paper subsequently carried a published correction.

Researchers working on these measures are notably more cautious than the companies selling them. The prevailing assessment is that biological age tests are valuable for research and not yet reliable enough to tell an individual consumer something actionable about themselves.

What DNA Methylation Actually Is

Methylation is a chemical modification in which a methyl group attaches to DNA, typically at sites where a cytosine sits next to a guanine. It does not alter the genetic sequence. It changes how accessible a stretch of DNA is to the machinery that reads it, influencing whether nearby genes are expressed.

This is a normal, essential regulatory system rather than damage. It is how a liver cell and a neuron, carrying identical DNA, become and remain different kinds of cell.

What makes it useful for measuring age is that methylation at particular sites changes in consistent, predictable directions across the lifespan. An epigenetic clock is a statistical model: measure methylation at a specific set of sites, weight them, and the output correlates strongly with chronological age. The interesting information is in the discrepancy between that prediction and the person's actual age.

Why There Are Dozens of Clocks, Not One

The plural matters more than any individual result. Different clocks were built by training on different target variables, which means they are answering genuinely different questions despite all reporting something called biological age.

The 2013 Horvath clock was trained to predict chronological age across multiple tissue types, and it does so with remarkable accuracy. Later clocks abandoned that target deliberately, on the reasoning that a model perfectly predicting chronological age carries no information beyond the birth certificate.

One recent unbiased comparison evaluated 14 separate epigenetic clocks against 174 incident disease outcomes. That study design exists precisely because the field has no agreed answer about which clock to use — and the clocks rank differently depending on which disease you care about.

What GrimAge Is Actually Trained On

GrimAge, frequently called the death clock, is the strongest mortality predictor among epigenetic clocks. Understanding why involves a detail that reframes several popular claims about it.

It was not trained to predict age. It was trained on DNA methylation proxies for plasma protein levels and for smoking pack-years, then optimised against mortality. It is an outcome-optimised model that happens to be expressed in units of years.

This is why smoking shows up so dramatically in GrimAge results. Smoking exposure is part of what the model was built to detect, so a heavy smoker receiving a substantially elevated GrimAge is partly the clock reporting back a variable it was explicitly designed to capture. That makes it a good mortality predictor and a poor demonstration that smoking 'ages you' by some independently measured number of years.

How DunedinPACE Differs

DunedinPACE takes a different approach again, and it is conceptually the cleanest of the three. Rather than estimating a static biological age, it estimates the rate at which someone is currently aging.

A score of 1.0 means aging at exactly one biological year per chronological year. Above 1.0 is accelerated, below is slower than expected. It emerged from a long-running cohort study that tracked participants' physiological decline over years, which is why it can express a pace rather than a snapshot.

In validation datasets it showed high test-retest reliability and predicted morbidity, disability and mortality with effect sizes similar to GrimAge, while adding predictive value beyond GrimAge alone. It is also the measure most suited to intervention studies, because a pace can plausibly change within a study window in a way a cumulative age cannot.

Where the 3.2-Year Reversal Number Comes From

Almost every claim about reversing biological age through lifestyle traces to a single 2021 pilot randomised trial, and its details are worth stating in full because they are consistently omitted.

Why the Within-Group Result Matters More

The trial reported a second number that receives almost no attention and is arguably more informative than the headline.

Compared against their own baseline measurements, treated participants showed a decrease of 1.96 years — with a p-value of 0.066, described in the paper as a strong trend towards significance. By conventional standards that result is not statistically significant.

The two figures differ because the 3.23-year result is a comparison between groups, so it reflects both the treatment group improving and the control group aging normally over the same period. Which framing you emphasise substantially changes the story, and the version that reached the public was the larger one.

The Clocks Don't Agree About You

For anyone considering paying for a test, this is the practical problem. Different epigenetic clocks do not always agree with one another even when applied to the same person's sample.

That is not a malfunction. Clocks trained on different outcomes are measuring different things, so divergence is expected. It does mean that a single biological age figure, presented without specifying which clock produced it and what that clock was trained to predict, conveys much less than it appears to.

There is currently no gold standard for measuring biological age, and researchers continue to work with numerous competing indicators. A consumer report saying you are 'biologically 38' is one model's output, not a measurement of a settled physical quantity.

What Changes Your Result Besides Aging

Reliability studies have found that technical and short-term biological factors can move results enough to swamp the effect someone is hoping to detect.

For standard replicate assays on EPIC and 450K arrays, nearly all clocks achieved excellent technical reproducibility — but some showed dramatic drops in reliability depending on where a sample sat on the processing slide and which DNA extraction protocol was used.

Why Researchers Say These Aren't Consumer Tests

The gap between the research literature and the consumer market is unusually wide here, and the researchers themselves are the ones pointing it out.

The assessment repeated across recent reviews is that biological age tests reveal what slows or hastens aging at population level while remaining useful for researchers rather than consumers. Important questions persist about reliability in commercial wellness settings and about how these measures should be used in clinical practice at all.

The distinction is between population inference and individual guidance. A clock that reliably shows a cohort of smokers aging faster than non-smokers is scientifically valuable and tells you very little about whether your own personal result reflects your health, your lab's handling of your sample, or which model the company happened to license.

Conclusion

Epigenetic clocks are real, useful and genuinely interesting. Methylation patterns do track aging, GrimAge does predict mortality, and DunedinPACE does capture a meaningful pace of physiological decline. None of that is in dispute.

What does not survive scrutiny is the idea that a single number tells you your true age. There are dozens of clocks trained on different outcomes, they disagree about the same person, no gold standard exists, and results shift with lab handling and time of day. The celebrated three-year reversal came from 43 men over eight weeks, with a non-significant within-group result and a subsequent correction.

The useful version of this idea needs no test at all. The interventions that showed up in the trial — better diet, sleep, exercise and stress management — are the same ones supported by decades of far larger studies using ordinary health outcomes. You do not need to know your methylation age to know those are worth doing.

This article summarises published research for general information. It is not medical advice, and biological age testing should not be used to make health decisions without qualified clinical guidance.

Data Source and Attribution

Nature CommunicationsAging (pilot trial)Epigenetic Clock Reliability

Trial details, effect sizes and p-values come from the 2021 pilot randomised clinical trial on diet and lifestyle intervention published in Aging, together with its published correction. Clock comparison figures come from an unbiased comparison of 14 epigenetic clocks against 174 incident disease outcomes published in Nature Communications. Reliability findings come from published work on the biological versus technical reliability of epigenetic clocks, and clock construction details reflect the primary literature on Horvath, GrimAge and DunedinPACE.

FactsFigs reviews, cleans, and cross-checks every source dataset before shaping it into a data story. Each visualization is created and designed in FactsFigs Design Studio — an internal tool developed and owned by FactsFigs — and is the original work of a FactsFigs author, not an AI-generated copy of any existing graphic. Individual assets within a visual may or may not be produced with AI tools, but the design of the visual itself is solely FactsFigs' own.

This content is for information only and is not medical advice, diagnosis, or treatment guidance. Figures reflect published research at the time of writing.

2026-07-20