Biological age tests can disagree with themselves by years

Key takeaways

• In a 2022 Nature Aging study, six well-known epigenetic clocks gave readings up to 9 years apart when the same blood sample was run more than once.
• Different clocks are built to predict different things, so one person can get several “biological ages” that don’t match.
• In the two-year CALERIE trial, eating less slowed one clock, DunedinPACE, by about 2% to 3%, while two other clocks didn’t budge.
• No direct-to-consumer biological age test is FDA-approved as a diagnostic, and prices run from about $30 to more than $1,000.
• Treat a result as a rough sketch: never change medicines or skip care because of it, and judge your health by the measures your clinician already tracks.

The results of a home telomere test landed on one man in his 40s like a diagnosis. Telomeres are the protective caps on the ends of chromosomes, and the report claimed his matched an 80-year-old’s¹.

Convinced his time was short, he left his job and sold his home. He even delayed surgery on a precancerous skin spot, fearing an operation would wear his telomeres down further¹².

Mary Armanios, an oncologist and clinical director of the telomere center at Johns Hopkins, described the case in 2018 as a warning¹. A lab number about aging can cause real harm when people read it as a verdict.

Telomere kits were the first wave. Today’s popular products are epigenetic clocks, which estimate “biological age” from chemical tags on your DNA. Dozens of companies sell them, for roughly $30 to more than $1,000⁴.

The science behind these clocks is real, and some of it is remarkable. But the research also shows how much a single consumer result can wobble. Here is what the tests measure, where they fall short, and how to decide whether one is worth your money.

Your DNA carries chemical tags that shift with age

Your genes don’t change much over a lifetime. The way your cells use them does. One sign of that is DNA methylation: small chemical tags that sit on DNA and help switch genes on or off.

At certain spots, methylation changes in predictable ways as people get older. In 2013, two research teams turned that pattern into “clocks.” Steve Horvath’s appeared in the journal Genome Biology, and Hannum and colleagues published theirs in Molecular Cell³⁷.

Those first clocks were built to guess a person’s calendar age from a sample⁷. Later clocks aimed higher. GrimAge, for example, was designed to predict death and disease risk rather than birthdays³⁷.

A third type measures speed instead of total mileage. DunedinPACE comes from a New Zealand study of 1,037 people born in 1972 and 1973⁷. Researchers tracked 19 signs of how well their organs worked at ages 26, 32, 38 and 45. Then they built a blood test that estimates how fast each person was declining⁷.

A DunedinPACE score of 1.0 means about one year of biological aging per calendar year. In the Dunedin group, the slowest person aged 0.40 years per year, and the fastest aged 2.44⁷.

Clock type	Examples	Built to predict	How to read the number
First generation	Horvath (2013), Hannum (2013)	Calendar age7	An age in years; the gap from your real age is the “result”
Risk-based	GrimAge	Death and disease risk37	An age in years tied to risk, not birthdays
Pace of aging	DunedinPACE	Speed of decline over about two decades7	Years of aging per year; 1.0 is average

One more catch: consumer kits read blood or saliva, not your organs. Stanford neurologist Tony Wyss-Coray notes that a blood-based test mostly reflects the age of your blood cells, because it can’t reach DNA in other organs³.

The same sample can come back years apart

In 2022, researchers led by Albert Higgins-Chen and Morgan Levine at Yale checked how steady these clocks are. They ran the same blood samples more than once. For six prominent clocks, readings differed by up to 9 years between repeat runs⁵.

That’s technical noise: small measurement errors at the many DNA spots a clock reads. In one example the team gave, a clock put someone at 50 on one test and 59 on the next⁶¹².

The team found a fix. They used principal component analysis, a statistical method that pulls the strongest shared signal out of noisy data, to rebuild the six clocks. The new versions agreed within 1.5 years for most repeat samples⁵.

Not every company uses such methods. Stanford epidemiologist David Rehkopf estimates that many clocks carry roughly three years of error. A result saying you’re three years older, he says, could just as easily mean you’re three years younger³.

Matt Kaeberlein described something more striking in a March 2026 opinion essay on the website Lifespan.io. He wrote that the same sample, sent to one company, came back with estimates nearly 15 years apart¹¹. He calls it his own experiment, shared in a video rather than a published study, so treat it as an anecdote, not data¹¹.

The sample itself matters too. Biobehavioral health scientists Idan Shalev and Abner Apsley wrote in The Washington Post in April 2026 that diet, illness, time of day and sample type can shift results. Saliva and blood can give quite different estimates⁴. Stanford’s report adds that blood cell levels can change within a single day³.

Clocks measure different things, so they disagree

Even with perfect lab work, clocks don’t agree with each other. They were trained on different groups to predict different outcomes.

The DunedinPACE team checked how closely their measure tracked older clocks in a large British study. On a scale where 1.0 is a perfect match, the correlation with Horvath’s clock was just 0.13. With GrimAge, it was 0.58⁷.

Stanford’s report offers a picture of how this plays out. Imagine a 52-year-old whose wearable says 45 and whose blood test says 54³. Each number comes from different measurements, and none is the single “true” age.

Clocks can also move in surprising directions. Stanford geneticist Anne Brunet points to a debate over rapamycin, a drug that reliably extends life in animal studies. In people taking it at the doses described, epigenetic clocks read slightly older, not younger. To her, that shows scientists don’t yet fully understand what the clocks capture³.

There’s also the question of who the clocks were built on. The DunedinPACE authors note that most groups used to test it were of white European descent, so how well it works in other populations isn’t yet established⁷.

Eating less slowed one clock, a little

The strongest human test of a lifestyle change comes from a trial called CALERIE. It randomly assigned 220 healthy, non-obese adults at three U.S. sites to cut calories or eat normally for two years⁹. Randomly assigned means chance, not choice, decided who went into each group.

Participants were asked to cut calories by 25%. On average, they managed 11.9%⁸.

A 2023 Nature Aging analysis, funded by the National Institute on Aging, looked at 197 of them⁸⁹. DunedinPACE slowed by about 2% to 3% in the calorie-cutting group compared with controls⁸. Two other clocks, PhenoAge and GrimAge, run in their noise-reduced versions, showed no significant change⁸.

The researchers noted that in other studies, a pace that much slower has been tied to a 10% to 15% lower risk of death, similar to quitting smoking⁹. That is an estimate borrowed from other research. The trial didn’t follow people long enough to see whether they got sick less or lived longer⁸.

For any one person, the effect was small. By the authors’ estimate, the spread of DunedinPACE results in the dieters and the controls was nearly the same, with close to 90% overlap⁸. The authors also caution that calorie restriction probably isn’t right for everyone⁹.

“Age reversal” headlines come from very small studies

In 2019, a study called TRIIM made headlines. Nine healthy men aged 51 to 65 took growth hormone for a year, plus DHEA and metformin to offset growth hormone’s tendency to promote diabetes¹⁰. On four epigenetic clocks, their biological age dropped by an average of 2.5 years¹⁰.

The study had no comparison group. Wolfgang Wagner of RWTH Aachen University in Germany told Nature the results weren’t solid, because the trial was very small and not well controlled¹⁰. Even Horvath, who ran the clock analysis, said he had expected slowing, not reversal¹⁰.

Here is a reasonable inference from the noise research. When older clocks can swing by years on repeat runs, a few-year change in nine people needs a comparison group before anyone calls it reversal⁵¹⁰.

Telomere tests have a wider margin of error

Telomeres shorten as cells divide, which made them an early favorite for measuring age. Stanford’s experts say telomere length may not be the most accurate aging marker³.

Measurement is part of the problem. Armanios says the lab method many consumer companies rely on, quantitative PCR, can vary by about 20%, so a retest on another day may not match. The flow-FISH method her clinic uses varies by about 5%¹.

Biology adds a twist. Normal telomere length varies widely between healthy people, and within that range it can’t pinpoint anyone’s biological age². Unusually long telomeres aren’t a prize, either. They may be linked to higher cancer risk¹².

Telomere testing does have real medical uses. It helps doctors choose gentler transplant care, with less immune suppression, for some patients with short telomeres who have an inherited form of the bone marrow disease aplastic anemia or the scarring lung disease pulmonary fibrosis¹. In her view, the test belongs in the clinic, not as “molecular palm reading”¹.

What your body already tells you

People drawn to natural health often want to measure whether their choices are working. That instinct is sound. And clock research has mostly backed habits this community already values.

Stanford’s experts name a healthy diet, aerobic and strength exercise, sleep, stress management and close social ties as influences on biological age³. Population studies using clocks also link eating less, regular exercise, a healthy diet, enough sleep and not smoking with slower aging on average⁴.

Rehkopf adds a humbling finding. How healthy people say they feel tracks fairly well with whether they test biologically older or younger³. Your energy, strength, mood and sleep are information too.

The World Health Organization has a name for that kind of everyday function: intrinsic capacity, covering cognition, mobility, energy and mood³. You can notice those things yourself, and a clinician can check them.

Many of these products also lean on familiar numbers. Stanford’s report notes that some wearables and blood panels base their “biological age” on heart rate, cholesterol and blood sugar, markers doctors already use³.

Before you buy: a seven-question checklist

1. What will I do differently with the answer? If the answer is “nothing,” save your money.
2. Which clock is it? Ask whether it estimates calendar age, disease risk or pace of aging, and what it was built to predict⁷.
3. Has the company published repeat-test results? Ask how far apart two runs of the same sample land. Noise-reduced versions of popular clocks exist⁵.
4. Blood or saliva? Don’t compare results across sample types or companies⁴.
5. Am I testing on a normal day? Illness, diet and time of day can shift results⁴. If you retest, do it at the same time of day, when you’re well.
6. Am I retesting too soon? Clocks respond slowly to change³, so a retest a few weeks after a new habit is unlikely to tell you much.
7. Who sees my data? Read the privacy policy, especially anything about insurers⁴.

If you already have a result:

• Don’t start, stop or change any medicine because of it.
• Don’t panic over an “older” number, and don’t relax your screenings because of a “younger” one³.
• Take real concerns to your clinician, who can check blood pressure, blood sugar, cholesterol and other measures with known meaning.

What the evidence says

Claim	What research shows	Evidence level
Epigenetic clocks track aging and death risk across large groups	DunedinPACE predicted death and disease in long-running studies, with effect sizes similar to GrimAge’s7	Strong evidence, for groups
One consumer test gives your precise biological age	Repeat runs differed by up to 9 years on six clocks; many carry about 3 years of error35	Evidence says no
All clocks give the same answer	Agreement between clocks ranged from weak to moderate7	Evidence says no
Eating less slows aging in people	CALERIE slowed DunedinPACE 2% to 3%; two other clocks didn’t change; no disease data yet8	Promising but early
Drug combinations can reverse your biological age	One trial of nine men with no comparison group10	Not enough evidence
A telomere test reveals your biological age	Normal range is wide; consumer methods vary about 20%12	Evidence says no
A clock can guide your personal treatment choices	No consumer clock is FDA-approved as a diagnostic3	Not enough evidence

Questions readers ask

The bottom line

Epigenetic clocks are a real scientific advance. They help researchers see whether a diet or drug nudges the pace of aging across hundreds of people. They aren’t yet precise enough to tell one person, on one day, how old their body is.

If you buy a test, treat the number as a rough sketch, not a diagnosis. Your daily energy, strength and sleep, plus the measures your clinician already tracks, remain better guides to how you’re aging.

Keep exploring: interpreting wearable data and consumer genetic test accuracy.

Related reading

• When the company reading your labs also sells the fix
• What a wellness panel really costs after you get the results
• What a medical test can and cannot tell you
• Your wearable’s best number is your own normal, not the average
• DNA kits read a sliver of your genes, and the data can be sold

Sources

1. Vanchieri C. “At-home telomere testing is not a reliable marker of aging, researcher says.” Science News, 2018. sciencenews.org
2. Sawyer A. “Measuring the length of your telomeres can provide an insight to your biological age: fact or fiction?” BioTechniques, n.d. biotechniques.com
3. Williams S. “What biological age tests can – and can’t – tell you.” Stanford Report (originally Stanford Medicine), 2026. news.stanford.edu
4. Advisory Board. “What do ‘biological age’ tests really tell you?” Daily Briefing, summarizing Shalev I and Apsley A in The Washington Post, 2026. advisory.com
5. Higgins-Chen AT, Thrush KL, Wang Y, et al. “A computational solution for bolstering reliability of epigenetic clocks: implications for clinical trials and longitudinal tracking.” Nature Aging, 2022 (eScholarship record). pubmed.ncbi.nlm.nih.gov
6. Yale School of Medicine. “A computational solution for bolstering reliability of epigenetic clocks: implications for clinical trials and longitudinal tracking.” Yale News, 2022. medicine.yale.edu
7. Belsky DW, Caspi A, Corcoran DL, et al. “DunedinPACE, a DNA methylation biomarker of the pace of aging.” eLife, 2022. moffittcaspi.trinity.duke.edu
8. Waziry R, et al. “Effect of long-term caloric restriction on DNA methylation measures of biological aging in healthy adults from the CALERIE trial.” Nature Aging, 2023. nature.com
9. Columbia University Mailman School of Public Health, via ScienceDaily. “Calorie restriction slows pace of aging in healthy adults.” ScienceDaily, 2023. sciencedaily.com
10. Abbott A. “Trial hints at age clock reversal.” Nature, 2019. media.nature.com
11. Kaeberlein M. “When Doctors Prescribe Horoscopes: The Trouble With Biological Age Tests.” Lifespan.io (opinion), 2026. lifespan.io
12. Karolinska Institutet. “Improved reliability of epigenetic clocks.” KI News, 2022. news.ki.se