Biological age tests: what epigenetic clocks can and can't tell you

A biological age test estimates how old your body behaves, usually by reading DNA methylation patterns in blood or saliva and running them through an algorithm called an epigenetic clock. The best clocks predict disease and mortality risk across large populations, but a single result for one person is noisy: older clocks can disagree with themselves by years on the same sample. Treat the number as a directional signal to discuss with a physician, not a score to optimize.

That is the short answer. The longer one matters if you have already paid for one of these tests, or are about to, and are trying to work out what a "biological age of 38" means for a 44-year-old running a company.

What is an epigenetic clock?

Your DNA sequence stays mostly fixed for life. What changes is how it is packaged and read. One of the best-studied layers of that packaging is DNA methylation: small chemical tags attached at specific sites on the genome. Some of those sites gain or lose methylation in a consistent way as people age.

An epigenetic clock is a statistical model trained on thousands of samples to find the combination of methylation sites that best tracks some target. The target is the important part, because it defines what the clock is actually telling you.

Three generations of clocks

First-generation clocks were trained to predict chronological age. The best known is the multi-tissue clock published by Steve Horvath in 2013, alongside a blood-based clock from Gregory Hannum's group the same year. If your "DNA methylation age" on one of these is higher than your birthday age, the idea is that your tissues look older than average for your years.

Second-generation clocks, such as PhenoAge and GrimAge, were trained on health outcomes: blood markers of physiological decline, smoking exposure and mortality. They trade some accuracy on birthday age for better prediction of who gets sick.

Third-generation clocks measure pace. The main example, DunedinPACE, does not estimate how old you are. It estimates how fast you are aging right now, which is closer to what most founders actually want to know.

Consumer tests often report one of these, or a proprietary blend, as a single headline number. Before you read your result, find out which clock produced it. A "biological age" from a first-generation clock and a pace score from DunedinPACE are answering different questions.

How DunedinPACE was built

DunedinPACE comes from one of the longest-running studies of human development. The Dunedin Study in New Zealand has followed roughly 1,000 people born in 1972 and 1973 since birth. Researchers led by Daniel Belsky at Columbia, with colleagues at Duke, Otago, King's College London and Harvard, used measurements taken at ages 26, 32, 38 and 45.

At each visit, the team tracked 19 biomarkers of organ-system integrity. The 2022 paper in eLife describes them as covering "cardiovascular, metabolic, renal, hepatic, immune, dental, and pulmonary systems." Over 20 years, some people's numbers deteriorated faster than others. The researchers then trained a methylation algorithm, from a single blood draw, to predict that rate of decline.

The output is scaled so that 1.0 represents one year of biological aging per calendar year. A result of 1.1 suggests you are aging about 10 percent faster than the reference average, and 0.9 suggests about 10 percent slower.

Does it predict anything that matters?

In the eLife paper, the team tested DunedinPACE in other cohorts of older adults. In the Framingham Heart Study Offspring cohort (2,471 people), faster DunedinPACE was associated with higher mortality risk, with a hazard ratio of 1.65. In the VA Normative Aging Study (771 older men), the hazard ratio was 1.26. Faster pace was also linked to cardiovascular disease, stroke and limitations in activities of daily living.

The Columbia summary puts it directly: "Midlife and older adults with faster DunedinPACE were at increased risk for incident chronic disease, disability, and mortality." The researchers also report that it adds predictive value over and above existing measures.

That is a meaningful result. It is also a population result. A hazard ratio tells you how risk differs across groups on average. It does not tell you what will happen to you.

Why the same test can give you two different answers

Here is the part the marketing tends to skip. Methylation is measured on laboratory arrays, and every measurement carries technical noise. For a population study with thousands of people, that noise averages out. For one person looking at one number, it does not.

In 2022, a team at Yale led by Albert Higgins-Chen and Morgan Levine published a study in Nature Aging on exactly this problem. They found that estimates from commonly used clocks "can often differ by up to nine years" when two or more measurements are run on the same blood sample.

Sit with that for a second. You could send a tube of blood to a lab, have it measured twice, and get results that suggest you are either a fit 40 or a tired 49. Nothing about your body changed. Only the measurement did.

The Yale team's fix was a statistical technique called principal component analysis, which kept the aging signal while stripping out much of the noise. With their approach, "for most tests, two measurements done at the same time will differ by less than a year." That is a large improvement, and newer clocks like DunedinPACE were designed with reliability in mind from the start.

The practical lesson in that chart: reliability drops when the measurement platform changes. If you test once with one company and again a year later with another, you are not comparing like with like. Even with the same company, a lab switching array versions can shift your number.

What biological age tests can tell you

Used carefully, a good epigenetic clock can do a few useful things.

It can confirm a direction. If your pace score is well above 1.0 and your blood pressure, fasting glucose, waist size and sleep are also drifting the wrong way, the test is adding weight to a story your basic labs were already telling. That can be the push some founders need to take the basics seriously.

It can track research interventions. Clocks like DunedinPACE are increasingly used as outcomes in trials of diet, exercise and drugs, because they can register change faster than waiting years for disease to appear. That is where the science is most useful right now: comparing groups of people under controlled conditions.

It can prompt a better conversation. A physician who sees an unusual result may look harder at cardiovascular risk, metabolic health or inflammation. The clock is the prompt. The follow-up tests are the substance.

What they can't tell you

They cannot diagnose anything. No consumer biological age test is FDA-approved to diagnose disease or predict individual outcomes. Most are sold as wellness products or lab-developed tests, which means the claims on the report have not been reviewed the way a diagnostic would be.

They cannot tell you which lever to pull. A number that says "aging fast" does not say whether the cause is sleep, alcohol, blood pressure, an undiagnosed condition or measurement noise. You still have to look at the inputs.

They are weak at tracking small changes in one person. If the measurement can move by years on the same sample with older clocks, a two-year "improvement" after three months of a new routine may be real, or may be the lab. Newer clocks narrow this, but the uncertainty for an individual is still larger than a single decimal suggests.

They reflect the populations they were trained on. The Dunedin cohort is one birth cohort from one country. Validation in other cohorts helps, but it is fair to ask how well any clock generalizes to your particular background.

What the evidence says, and what it doesn't

Strong: DNA methylation changes with age in consistent, measurable ways. Epigenetic clocks track chronological age closely across large groups, and the better second- and third-generation clocks predict disease and mortality in large, independent cohorts.

Moderate: DunedinPACE was built on 20 years of longitudinal biomarker data in a well-characterized cohort and shows excellent test-retest reliability on the same platform. Its predictions for disease, disability and death have been replicated in several cohorts of older adults. This evidence is observational: it shows association, not that changing your score changes your outcome.

Early or mixed: Whether lifestyle changes that move a clock score also change real health outcomes for an individual is not established. Trials using clocks as outcomes are growing, but most are short and small.

Not supported: Any claim that a supplement, peptide or protocol "reverses your biological age by X years" based on a clock result. Given the known measurement noise, those claims need controlled trials with real outcomes before they mean much.

How a founder should actually use a biological age test

If you are curious, testing is low-risk physically. The risk is spending money and attention on a number instead of on the things it is supposed to reflect. Here is a sane way to approach it.

Get the basics first. Before or alongside any clock, know your blood pressure, lipids, fasting glucose or HbA1c, waist measurement and a recent fitness marker. These are well understood, cheap and directly actionable. If you have not had a physical in two years, that is the higher-value appointment.

Ask which clock, which lab, which array. A reputable test should tell you which algorithm produced your number. If the report gives only a proprietary "biological age" with no method, discount it accordingly.

Retest with the same provider, and not too often. If you want a trend, keep the provider constant and space tests at least several months apart. Look for large, sustained shifts, not a one-year swing.

Read it next to your life, not instead of it. If your pace score came back fast during the worst quarter of your fundraise, when you were sleeping five hours and drinking most nights, that is a coherent story. The score did not reveal anything new. It put a number on what the calendar already showed.

Take surprises to a physician. If a result is alarming, or you have symptoms like chest pain, shortness of breath or unexplained weight change, see a physician rather than trying to fix it with a new routine alone.

Where stress fits

Chronic stress shows up in many of the systems DunedinPACE was built to track: blood pressure, metabolic markers, inflammation, sleep. We cover the physiology in detail in what chronic stress does to a founder's body. A clock result will not tell you how much of your number is stress. But for many founders, the honest inventory of inputs starts there, and with the short nights described in meditation for sleep for executives.

This is also where the most durable changes tend to come from: a nervous system that can come down after the board meeting instead of staying wired until 1 a.m.

Building the baseline the number is trying to measure

A biological age test is a snapshot of years of inputs. The inputs are where your leverage is: how you sleep, how you recover from pressure, whether your stress response actually turns off at night.

That is the work we do at Cinder: private 1:1 coaching that trains stress regulation and recovery as skills, fitted to a founder's real calendar, with sessions online and practice between them. It is not a biomarker program and makes no claims about your clock score. It is a way to change the daily inputs those scores are built from, and to keep them in place when the quarter gets hard.

Apply for private coaching

Sources

• DunedinPACE, a DNA methylation biomarker of the pace of aging, eLife, 2022
• Epidemiologists develop state-of-the-art tool for measuring pace of aging, Columbia University Mailman School of Public Health, 2022
• A computational solution for bolstering reliability of epigenetic clocks: implications for clinical trials and longitudinal tracking, Yale School of Medicine (Nature Aging study), 2022