A 2026 audit pooled 51 human trials and ran the same 16 epigenetic clocks across every one. Only 19 treatments did anything to slow biological aging. Just 13 still counted after the authors allowed for testing so many measures at once (Study). Diet and exercise programs did most of that work, along with a few prescription drugs. Supplements off the shelf mostly left the number where it was.
What an Epigenetic Clock Measures
Your DNA carries tiny chemical tags, and adding one is called DNA methylation. The pattern of those tags shifts in fairly predictable ways as you get older.
An epigenetic clock is a formula built on that pattern. It reads a few hundred tag sites, weighs each one, and gives back a single number in years.
One split runs through the rest of this article. First-generation clocks, such as Horvath and Hannum, were trained to guess your date of birth from your blood. Second-generation clocks were trained on something else: records of who got ill and who died. PhenoAge and GrimAge belong to that group. DunedinPACE is different again. It estimates speed rather than age, and asks how fast a body is wearing out right now.
So these are different rulers. They do not measure the same thing, and they do not have to agree.
Real data shows that plainly. A Scottish study tracked 18,859 adults over 10 years, and researchers ran 14 clocks against 174 diseases. The two first-generation clocks each flagged a single disease, while GrimAge and DunedinPACE flagged dozens (Study). Even so, the authors found no clear best clock overall.
Whether these tests can measure one person reliably is a separate question, and a serious one. Our piece on what a biological age test really tells you covers that side. This article asks the other half. What actually shifts the number?
What Can Slow Biological Aging
The new audit did something the field had never done. Instead of running one more trial, a Yale lab collected 51 trials that had already finished. Each one had blood samples from before and after. The team then recalculated the same 16 clocks in every dataset, plus 94 other methylation measures (Study). That is the point of the exercise. Every original paper had used its own clock and its own methods, so nobody could compare them.
Now the scorecard. Of everything tested, 19 treatments lowered epigenetic age, and 13 of those survived the correction for testing so many measures at once. Five pushed epigenetic age the other way, upwards (Study).
It is worth knowing how gentle that correction was. The authors picked a deliberately lenient method. A strict one, they argued, would erase real signals, because the clocks overlap so heavily. So 13 is the optimistic count, not the cautious one.
Two families of treatment did the work: prescription drugs and lifestyle change. Drugs produced the bigger average shifts, and lifestyle changes produced the more consistent ones (Study).
Two of the findings surprise almost everyone.
The first is that who was in the trial mattered more than what they took. Of the 51 studies, 41 ran in healthy people and 10 in people with a diagnosed illness. In the ill groups, eight different clocks dropped. In the healthy groups, only three moved at all, and by far less (Study).
The second is that longer studies saw more movement. Study length was one of the two biggest predictors of whether anything happened at all (Study).
Put those together and the conclusion is uncomfortable. A six-week plan in a healthy 40-year-old is close to the worst possible setup for finding a real effect.
The clocks split as expected. The ones trained on illness and death responded most, and so did the pace-of-aging clock. They also agreed with each other. The birthday-guessing clocks moved on and off, and their movement tracked random noise more than biology (Study).
One disclosure belongs here. Three coauthors work at TruDiagnostic, a company that sells epigenetic age tests, and it supplied private data for the database. The first and senior authors report consulting fees from it (Study). That is ordinary in this field. It is still worth knowing.
Food and Movement Won
The most useful finding is also the least exciting. Diet and exercise programs together lowered epigenetic age, and the type of diet barely mattered (Study). Mediterranean, low-carb, low-fat: the pairing did the work, not the argument about carbs versus fat.
CALERIE is the cleanest single test we have. A coin flip sorted 220 adults without obesity into two groups. Two thirds cut calories by about a quarter for 2 years, and the rest ate freely. A later analysis ran the clocks on their stored blood (Study). DunedinPACE, the pace-of-aging measure, slowed by roughly 2 to 3%. PhenoAge and GrimAge did not budge.
Read that twice. Inside one well-run trial, three clocks gave three different answers. The authors called the change small and said so in print. Most dieters never reached the cut they were asked for, and averaged closer to 12%. Our full write-up of what the CALERIE trial proved covers the rest of that story.
Movement has scale behind it, but weaker study designs. A 2026 review gathered 44 studies covering 145,465 adults (Meta-analysis). More physical activity went with younger readings on the Horvath and GrimAge clocks, while two other clocks showed nothing. Only 7 studies were comparable enough to pool, and all 7 were snapshots taken at a single moment.
A snapshot cannot tell you which way the arrow points. Maybe moving more keeps a body younger. Maybe a younger body finds it easier to move. The authors say this outright, and warn that the evidence cannot establish cause (Meta-analysis).
Exercise on its own has also come up empty in trials. One study followed 777 older adults for 3 years, and a home exercise program moved none of the four clocks tested (Trial).
So the honest version is this. Diet and movement together carry the most consistent clock evidence of any lifestyle change. Exercise tested on its own has been quieter. Neither result changes what you already knew about food and walking.
Drugs In, Supplements Out
The strongest responses in the whole database came from prescription drugs, and not new ones (Study). All were already prescribed for something else. Metformin, semaglutide and anti-TNF drugs all moved the clocks. Anti-TNF drugs damp down inflammation in conditions like rheumatoid arthritis and inflammatory bowel disease.
That anti-TNF result was the most consistent of the lot. Across separate studies in arthritis and bowel disease, nearly every second-generation clock shifted by similar amounts (Study).
Treat it as a clue about mechanism, not a shopping list. Metabolic strain and inflammation look like a large part of what these clocks are reading. Metformin’s shift, for example, landed mainly in the inflammation, brain and metabolic sub-scores (Study).
The trials themselves explain why this is not advice. The semaglutide study ran 32 weeks in 84 people, all being treated for an HIV-related fat condition. The metformin study ran 24 weeks in 11 people with HIV, and a second, longer metformin trial in 35 people found nothing at all (Review). Most of these trials ran in patient groups already under treatment. They were not healthy people taking a drug to slow aging. The few exceptions are tiny. One gave growth hormone plus DHEA and metformin to 9 healthy men for a year. Another gave senolytics to 19 healthy people for six months and changed nothing (Review).
Supplements tell the opposite story. The over-the-counter aisle, the one with the marketing budget, largely failed to move the number.
Two exceptions are worth naming. The biggest clock analysis of a supplement trial followed 777 adults aged 70 and over for 3 years (Trial). One gram of omega-3 a day slowed three of the four clocks tested. Vitamin D alone did nothing, and the exercise program alone did nothing. All three together added up on PhenoAge only.
Then look at the size. Over 3 years, the whole shift came to about 3 to 4 months of biological age. That is a real change, but it is nowhere near rejuvenation. The clock finding came from stored blood analyzed after the fact. The trial’s six main health outcomes had all come back empty (Trial).
A separate 2-year trial in 958 adults found the same shape with a daily multivitamin. Two of five measures shifted by a small amount, and the other three showed no clear change (Trial). The benefit sat almost entirely in the people who were already aging fastest.
The supplements that failed share a feature. They were small and short: 19 people for 30 days, 16 people for 8 weeks, 20 people for 10 weeks (Review). By the audit’s own logic, those studies were built not to find anything.
Clock Change Is Not Lifespan
Everything above rests on an assumption nobody has tested yet.
No trial has shown that lowering an epigenetic clock leads to less disease or a longer life. The clocks do predict death well when they are measured once across a large group. That is a different claim from saying a change in your number changes your future.
The Biomarkers of Aging Consortium wrote the checklist for this, and its verdict is blunt. No aging marker is proven to stand in for a real health outcome (Review). Showing that a marker responds to treatment is step one, not the finish line. The authors note that stand-in measures have failed before in medicine.
Steve Horvath built the first epigenetic clock. On the day the audit appeared, he published a commentary in the same journal. He called it “Putting epigenetic aging clocks on trial” (Commentary). The clocks are the thing on trial now, not the treatments.
The audit’s own authors agree. They write that it is “still unclear whether short-term biomarker changes correspond to long-term improvements in disease, healthspan or lifespan” (Study). They also name the missing piece. Nobody has defined how big a clock change must be before it means anything for a patient.
It helps to see how thin the famous results are. TRIIM is the study people cite for “aging reversed”. It gave growth hormone, DHEA and metformin to 9 men for a year. There was no control group, no dummy pill and no coin flip (Study). The headline figure of 2.5 years younger was measured against a predicted untreated path, not against real people. The drop from the men’s own starting point was about 1.5 years.
Two teams can read this literature and disagree. A 2026 review by Johnson and Sinclair listed a far longer set of treatments that work. The same review found that rapamycin, nicotinamide riboside and senolytics did nothing, and that plasmapheresis nudged the clocks slightly faster (Review). That review is an inclusion list with no pooling and no quality scoring, and it counts TRIIM as a win.
One more thing to hold while reading any of it. The people who build these clocks often own a piece of them. One author of that review works for a company selling epigenetic age tests, and the other founded it (Review). Horvath declares his own stake in clocks he helped invent (Commentary). In this field, the people judging the tests often sell them too.
What a Mail-In Test Misses
The best treatments in the audit move a clock by a few months over several years. Set that beside the wobble in the measurement itself. Researchers split single blood samples in two and ran both halves. Across six popular clocks, the authors report gaps of up to 9 years between the two answers (Study). Even the typical gap ran to a year or two.
The authors did that arithmetic themselves. Say a treatment lowers epigenetic age by 2 years. Error of this size at both ends would bury the result.
Be fair about which clocks, though. The 9-year figure comes from first-generation clocks and PhenoAge. DunedinPACE was built to skip the noisiest tag sites, and its makers credit that for its steadier repeat readings (Study). So it is wrong to say that every clock is 9 years noisy. The point is that even the steadiest one carries error, and that error is large next to a change measured in months.
A single before-and-after on a mail-in kit therefore cannot tell you whether your diet change did anything. A reading that dropped 3 years is at least as likely to be the test as the treatment. There is a related trap here. When one clock out of several shows a change and the rest stay flat, it is usually the clock, not you (Study).
So why did the audit work? It had hundreds of people per study, control groups picked by coin flip, and the same clock computed the same way everywhere. Random error cancels out across a crowd. It does not cancel out across your two tubes.
Plenty is worth tracking instead, over a year, for almost nothing. Watch your resting heart rate, your waist size and your blood pressure. Notice how a hard uphill walk feels. Add fasting glucose or HbA1c, grip strength, and what time you fall asleep. Each one is cheap, repeatable, and already tied to outcomes that matter.
Frequently Asked Questions
Can you actually slow biological aging?
Some things do move the number. In the pooled audit of 51 trials, 19 treatments lowered epigenetic age, and 13 held up after correction (Study). But no trial has yet shown that moving a clock means less disease or a longer life (Review). So some treatments may slow biological aging as these tests measure it. That is not the same as slowing aging itself.
Do supplements lower your epigenetic age?
Mostly not, with two real exceptions. A 3-year trial in 777 older adults found that 1 gram of omega-3 a day slowed three clocks (Trial). A 2-year trial in 958 adults found a small shift from a daily multivitamin (Trial). Both effects were measured in months, not years. Most other supplement studies were small and short, and they found nothing (Review).
Is a $300 biological age test worth it?
Not for checking whether your own routine is working. Split one blood sample in two, and popular clocks can return ages years apart (Study). The best treatments shift them by months. Blood pressure, waist size, fasting glucose and grip strength cost far less and hold steadier.
Which epigenetic clock is the most accurate?
Second-generation clocks such as GrimAge and DunedinPACE respond better to treatment. They also agree with each other more than the birthday-guessing clocks do (Study). In a study of 18,859 adults they flagged many more diseases (Study). But no clock is proven to tell one person anything about their own future (Review).
Key Takeaways
- 19 lowered the clock, 13 held up. One audit recalculated 16 epigenetic clocks across 51 human trials, and most treatments did not move the number (Study).
- Food plus movement beat the supplement aisle. Diet and exercise together lowered epigenetic age whatever the diet, and CALERIE slowed one pace-of-aging measure over 2 years (Study).
- A few prescription drugs moved it, in patients. Metformin, semaglutide and anti-TNF drugs shifted the clocks in people already under treatment. That is not a reason to seek them out (Study).
- A lower clock has never been shown to mean a longer life. The consortium framework says no aging marker is proven to stand in for a real health outcome (Review).
- The noise is about the size of the prize. Re-running one blood sample can change the answer by years, while the best treatments move it by months (Study).
Start With the Boring Basics
There is a pattern in the scorecard that is easy to miss, because it is so unglamorous. The treatments with the best evidence were already on the list before anyone built a clock. Eat mostly real food. Move most days. Keep a regular sleep schedule. Treat the condition you actually have, since that is where the biggest shifts turned up.
You do not need a kit to begin. Pick one thing this week and give it a year: a 30-minute walk after your largest meal, most days. Then check the cheap markers in twelve months, starting with resting heart rate, waist size and blood pressure.
A clock ticking backwards is not the best part of this story. The better news is that a field built on selling numbers has started testing its own instruments in public. Checking your own tools in the open is how a measurement earns trust.
This article is for educational purposes and is not medical advice. Talk to a qualified clinician before changing your health regimen.
Leave a comment