A biological age test reads chemical tags on your DNA and hands back an age in years. Run one blood sample through the lab twice, and six popular clocks can disagree by as much as 9 years (Study). That gap is the problem. A biological age test is useful for research on groups. It is not steady enough to track one person over time. The science behind these clocks is careful work. The number printed on your PDF is shakier than it looks.
What a Biological Age Test Measures
Your DNA carries small chemical tags called methyl groups. They sit on the strand and help switch genes on and off. That pattern changes as you age, and it changes in fairly predictable ways.
An epigenetic clock is a formula built on that pattern. It reads a few hundred tag sites, weighs each one, and returns a single number in years.
There are three families, and they answer different questions. First-generation clocks were trained to guess your date of birth from your blood. Second-generation clocks were trained on illness and death records instead, so they aim at risk rather than birthdays. Pace-of-aging measures estimate speed. DunedinPACE is the best known of them. It came from tracking 19 body systems in the same people at ages 26, 32, 38 and 45 (Study). A score of 1.0 means one biological year per calendar year.
Which family you get matters more than the brand name. A clock trained to guess birthdays answers a question you can already answer yourself. A clock trained on illness records at least aims at something you care about.
What arrives in the post is simpler than any of that. You get a saliva tube or a finger-prick card, a prepaid envelope, and a PDF a few weeks later. The PDF names an age. It rarely names which clock produced it. Some companies never publish their formula at all (Review).
Same Blood, Different Age
Picture a bathroom scale that reads 78 kg. You step off, step straight back on, and it reads 84 kg. Nothing about you changed. The scale is just unsteady. Scientists call that test-retest error.
Researchers ran that experiment on epigenetic clocks. They measured the same blood samples twice, as a straight repeat. Same tube, same day, no biology in between. Across six well-known clocks, the two answers landed as much as 9 years apart (Study).
That was the worst case rather than the usual one. The typical gap between two runs ran from about 1 to 2 and a half years. Now set that against the biology. Between different people, the whole spread in this measure is only 3 to 5 years. So the biggest errors were as large as the differences the clocks are built to spot.
The same team built a fix. They rebuilt all six clocks on principal components, a way of pooling many tag sites so random blips cancel out. More than 9 in 10 repeat pairs then agreed within about 1.5 years. The typical gap fell to well under a year (Study).
One measure came out of this well. DunedinPACE gives nearly the same score when a duplicate blood sample is run, and it repeated better than most clocks it was tested against (Study).
So a report saying you are 4 years younger than your birthday tells you nothing on its own. The ruler alone could have produced that gap. The fix exists, and researchers can use it for free. Your consumer report rarely says whether it was used.
The Number Moves All Day
Lab error is only half the problem. The reading also drifts while your biology sits still.
Two healthy men in their thirties gave blood 24 times over 3 months. Nobody changed anything in their lives on purpose. Their epigenetic age wandered anyway. From one collection day to the next it moved by more than 3 years, and by more than 6 years on one clock (Study). Correcting for the mix of cells in the blood did not remove the drift. Two men is a tiny study, and the authors say so themselves.
The reading also swings inside a single day. One healthy 52-year-old man had blood drawn every 3 hours for 3 days. Thirteen of the 17 clocks tested showed a daily rhythm. Readings came out youngest around midnight and oldest around midday, by up to about 5.5 years on one clock (Study). That finding rests on one man, so hold it loosely.
Then there is the tube itself. Researchers sampled 83 people across five tissues on the same day. Cheek and saliva samples returned very different ages from blood samples. On some clocks the gap inside one person averaged close to 30 years (Study). On the weakest pairings, a saliva result told you almost nothing about that person’s blood result. Blood samples matched other blood samples closely. These clocks were built on blood.
Different formulas disagree too. In the Dunedin study, 818 adults had their DNA tags measured at age 38. Three epigenetic clocks run on those same samples agreed with each other only loosely (Study). So two companies can hand the same person two different ages, with both doing their job correctly.
One number from one tube is one noisy reading. It is not a verdict on how you are aging.
Where the Clocks Earn Their Keep
None of that makes the clocks junk. Across large groups of people, they work.
Researchers pooled 13 studies covering 13,089 adults. People whose clock ran ahead of their birthday died sooner more often. The link held after allowing for smoking, weight, blood pressure, diabetes, exercise, education and alcohol (Meta-analysis). A second pooled analysis of nearly 18,000 people found the same signal for early death (Meta-analysis). Its authors warn that the disappointing studies may never have been published. They also say the link to specific diseases is still unclear.
Not every clock does this equally well. In a national US sample of 2,105 adults over 50, the newer GrimAge measure predicted death most strongly. It came out ahead of the older Horvath and Hannum clocks, and ahead of PhenoAge (Study).
The limits show up group by group. In that same sample, several clocks predicted death in non-Hispanic white participants but not in Hispanic participants (Study). A tool that works on average can still miss the person in front of it.
Research is where these tools do their best work. The CALERIE trial used a coin flip to sort 220 adults into two groups. One group cut calories by about a quarter for 2 years, and the other ate normally. DunedinPACE slowed in the calorie-cutting group by about 2 to 3%, at 12 months and again at 24 (Trial). The static age clocks stayed put. In practice the dieters managed roughly half the cut they were asked for. Our write-up of what the CALERIE trial proved covers the eating side of that story.
A 2026 trial did the same thing with a daily multivitamin. It followed 958 adults for 2 years and used five aging measures, four of them principal-component clocks (Trial). Two of the five shifted. One independent expert put the size of the shift at about 2.7 to 5.1 months against a dummy pill. Another said this is not convincing evidence that a daily multivitamin meaningfully slows aging (Commentary).
That split is not a one-off. A 2026 project gathered 51 studies of treatments and 3,128 samples, then ran the same 16 clocks over all of them (Analysis). Pace-of-aging clocks and death-trained clocks responded most consistently. The first-generation clocks moved here and there, with no clear pattern.
The same principle sits behind all of it. Average across hundreds of people and the random wobble cancels out, leaving the real signal. One person gets no such cancelling.
Why It Fails as a Dashboard
The noise is bigger than the effect. A real result from a real treatment needed 958 people and 2 years to show up. It came to a few months (Trial). One person’s repeat test can move by years (Study). Even the rebuilt clocks can differ by up to about a year and a half between two runs. That is several times the multivitamin effect, produced by noise alone.
No study has shown that a clock can tell one person whether a supplement, a peptide or a protocol is working. A review by the Biomarkers of Aging Consortium is blunt about it. No regulator has signed off on any aging marker as a stand-in for aging. Nobody has even agreed on how to prove that one works (Review). Consumer kits sit in a general-wellness lane that the FDA does not review before sale (FDA). A 2026 review of the commercial products lands in the same place. They are not reliable enough to guide one person’s decisions (Review).
Then there is the retest trap, which is what sells subscriptions. You test, get an unflattering number, change something, and test again. Extreme readings tend to be followed by less extreme ones, purely by chance. Statisticians call that regression to the mean. Add the lab noise on top. A second test reading younger is the most likely outcome whatever you did in between.
A frightening result has a cost too. A number saying you are 8 years older than your birthday arrives with no proven action attached. Even among the newer clocks that do respond to things, only about a third of the treatments tested moved them (Review). That tally comes from authors with their own commercial stake in the tests.
What to Track Instead
Plenty of numbers about your body are steady, cheap and worth moving.
Fitness is the strongest of them. One overview pooled 199 studies that followed people over time. Together they covered more than 20 million observations. Fitter people were far less likely to die from any cause, and far less likely to develop heart failure (Umbrella review). You can also train it, which is what makes it worth tracking. VO2max is the number behind it, and a timed walk will do if you have no lab nearby.
Blood pressure pays you back when you move it. Across 48 trials and nearly 345,000 people, lowering the top number by 5 points cut major heart events by about 10% (Meta-analysis).
Grip strength costs almost nothing to check. In just over half a million UK adults, a weaker grip went with a higher chance of dying. It also added information that a standard clinic risk score missed (Study).
Round it out with a tape measure at your waist and an ordinary blood panel, including HbA1c and cholesterol. Your doctor already reads those. Each one comes with an action attached.
Cost matters here as well. Most consumer kits run to hundreds of dollars, and a repeat subscription multiplies that.
If you still want to test, waste as little as possible. Prefer a pace-of-aging measure or a principal-component version, because they repeat far better (Study). Use the same lab and the same sample type every time. Blood and saliva are not interchangeable (Study). Book the draw at the same time of day (Study). And treat any change smaller than a few years as nothing at all.
Frequently Asked Questions
Are Biological Age Tests Accurate?
A biological age test is accurate enough for research on groups, and not accurate enough for one person. Running the same sample twice on six popular clocks produced answers up to 9 years apart (Study). Newer principal-component versions cut the typical gap to well under a year. The clocks do predict death across large populations (Meta-analysis). That is a different job from reading your own PDF.
Why Did Two Biological Age Tests Give Me Different Results?
Three reasons stack up. Different companies use different clocks, and different clocks agree only loosely on the same sample (Study). Saliva and blood are different tissues, and the gap inside one person can average close to 30 years (Study). On top of that, plain lab noise moves the answer by years (Study).
Can You Lower Your Biological Age?
At a group level, some things do slow these measures. Cutting calories for 2 years slowed DunedinPACE by about 2 to 3% (Trial). A daily multivitamin shifted two of five measures by a few months (Trial). Across the wider research, only about a third of the treatments tested moved even the responsive clocks (Review). None of this has been shown to turn up as a reliable change in your own report.
Is a Saliva Biological Age Test as Good as a Blood One?
Most clocks were built on blood, and they do not transfer cleanly. When 83 people gave both, saliva and cheek results sat far from blood results (Study). One clock, the skin-and-blood version, came closest across tissues. Even that one was not interchangeable. If you test at all, never compare a saliva result with a blood result.
Key Takeaways
- Repeat the same sample and the answer moves by years. Lab noise alone put two runs of one blood sample as much as 9 years apart on six popular clocks (Study).
- The reading drifts from day to day. Two men sampled 24 times over 3 months saw their epigenetic age swing by more than 3 years between collection days (Study).
- Pace-of-aging and rebuilt versions repeat far better. DunedinPACE gives nearly the same score on duplicate blood samples (Study), and more than 9 in 10 repeats of the rebuilt clocks agree within about 1.5 years (Study).
- They predict outcomes in populations, not in people. Pooled across 13 studies and 13,089 adults, a fast clock tracked earlier death (Meta-analysis). Yet several clocks failed to predict it in Hispanic participants in a US national sample (Study).
- Nothing shows that a clock works as a personal progress bar. No regulator has signed off on any aging marker as a stand-in for aging. Nobody has agreed on how to prove that one works (Review).
A Ruler, Not a Report Card
Wanting a number for how you are aging is a reasonable thing to want. The people building these clocks are doing serious work, and the field has earned real results with them. CALERIE and the multivitamin trial could not have been read out any other way.
The honest answer is narrow and a bit dull. The ruler is steady enough to compare a thousand people. It is not yet steady enough to compare you with yourself.
None of that means you should ignore your aging. It means measuring it well takes more than a tube in the post.
So keep an eye on the boring numbers instead. Your fitness, your blood pressure, your waist, your grip, your blood panel. They move when you work on them, and the movement means something. Nobody needs to sell you a subscription to read them.
The evidence on epigenetic clocks is still building, and it is worth watching. Just do not let one drive your health decisions yet.
This article is for educational purposes and is not medical advice. Talk to a qualified clinician before changing your health regimen.
Leave a comment