There's a test that claims to tell you how fast you're ageing. It costs a few hundred dollars, needs a cheek swab or a drop of blood, and a multi-million-dollar industry has been built on watching the number move. This week, researchers did something nobody had done before: they took the industry's own data and audited it. Most of the clocks barely budged. The supplements did nothing. The boring stuff — drugs, the Mediterranean diet, sleep — is what actually moves the needle. If you paid for a biological age, this is the study that tells you what you bought.

The number people pay to watch

Epigenetic clocks are real science. The idea, which traces to Steve Horvath's landmark 2013 clock, is that the pattern of chemical marks on your DNA — methylation — drifts in a broadly predictable way as you age. Measure the drift, and you get a number: your epigenetic age. And that number genuinely means something as a snapshot: people whose epigenetic age runs ahead of their chronological age die sooner. Cross-sectionally, the signal is real and robust. That part was never in dispute.

What consumers are actually buying is not a snapshot. It's a dial. The direct-to-consumer industry — the few-hundred-dollar test, the "reverse my age by X years" protocol — runs entirely on the change metric: the number going down after you take the product. That is the whole business model. And until this week, nobody had checked whether the change metric works.

What the audit actually did

On 21 August, Sehgal and colleagues published the first systematic audit of epigenetic clocks in Nature Medicine. They harmonised 51 intervention studies from the TranslAGE database — every human study they could find that measured epigenetic age before and after a longevity intervention — and ran 16 epigenetic clocks plus 94 other DNA methylation biomarkers across them.

The test they applied comes from clinical-trials logic that predates the longevity industry: Fleming and Powers's 2012 framework for biomarkers and surrogate endpoints. A surrogate earns its keep only if it does two things — it responds when patients genuinely benefit, and it stays put when they don't. The audit held every clock to exactly that standard.

And here is the detail that should make the industry wince. The data did not come only from public repositories like GEO and EMBL. It included private trial data from TruDiagnostic — one of the companies selling these tests to consumers. The industry was audited on its own books.

Most of the clocks don't clock

The headline result is uncomfortable for anyone who has ever posted a screenshot of their biological age going down. Of the 51 interventions, 19 significantly decreased epigenetic age — and only 13 survived correction for multiple testing. Five made ageing look worse. Twenty-six — more than half — did nothing at all.

The first-generation clocks — Horvath's original and the Hannum clock — were, in the paper's word, "sporadic": no clear trend in either direction. If your few-hundred-dollar test runs on one of those, you have been watching noise.

The clocks that passed are the mortality and pace-of-ageing clocks: GrimAgeV2, PCGrimAge, PCPhenoAge, DunedinPACE and SystemsAge — the ones that respond to real benefit and hold still when there is none. The standout was DunedinPACE, the pace-of-ageing measure developed by Belsky and colleagues in 2022: it decreased in 16 interventions and increased in only one. That is what an instrument looks like.

What actually moves the number

Who actually moves the clocks? Who actually moves the clocks? Mean effect on epigenetic age, by intervention category −0.093 −0.039 ≈0 (n.s.) ≈0 (n.s.) Pharmacological Lifestyle Supplements Medical procedures −0.10 −0.05 0 Mean change in epigenetic age (years — negative is younger) 19 of 51 interventions moved the clocks down · 26 did nothing · 5 made ageing look worse Source: Sehgal et al., Nature Medicine (2026); n.s. = not significant
Chart: mean effect on epigenetic age by intervention category, from the TranslAGE audit.

The effect sizes tell a story the marketing departments would rather you didn't read. Pharmacological interventions — drugs — moved the clocks down by 0.093 years on average: significant. Lifestyle interventions — diet, exercise — managed 0.039 years: also significant. Drugs beat lifestyle by a margin that was itself significant (P=0.0063), and they crushed supplements (P=0.0009). Supplements: effectively nothing. Medical procedures: effectively nothing.

The reproducibility picture is worse for the supplement aisle. Anti-TNF therapy and the Mediterranean diet moved the clocks consistently, across every study that tested them. Senolytics — the "clear your zombie cells" class that has sold a lot of capsules — were inconsistent across all five studies. And the two supplement trials run by SRW moved the clocks in opposite directions: some clocks down, others up, for the same product. When your own product cannot decide which way the dial turns, the dial is not broken. The product is.

The industry audited itself — and lost

Savour the symmetry. The audit's private data came from TruDiagnostic, which sells the tests. The most embarrassing supplement results came from SRW's own studies, which sell the pills. And the metric's inventor — Steve Horvath, named inventor on the GrimAge and PhenoAge patents, founder and consultant of the Epigenetic Clock Development Foundation, and until March a principal investigator at Altos Labs with equity in the enterprise — co-published the accompanying News & Views in the same issue — literally a report card for ageing-clock surrogacy, financial disclosures printed in full — calling for exactly this validation discipline.

Let that land. The man with the most to lose from a failing clock wrote the editorial saying clocks must be tested. That is not a scandal. That is a field growing up.

The fix is measurement, not censorship

Here is the science-positive part, and it is the part the cynics will miss. The correct response to a hype market is not to ban the tests. It is to make the tests honest — and that is precisely what this audit does. The clocks that passed are now credible trial endpoints. Interventions that genuinely move them — metformin and rapamycin-class drugs, anti-TNF, the Mediterranean diet — can be tested faster, cheaper and with smaller samples, because the outcome is a number instead of decades of follow-up. Validation science just shortened the road to answers.

For the consumer, the rule is almost embarrassingly simple. Stop paying to watch the number. If you're in a trial or buying a protocol, demand the clocks that passed the audit — mortality and pace-of-ageing measures like DunedinPACE, not the first-generation Horvath or Hannum dials. And start doing the things that actually move them: drugs, when a clinician prescribes them; the Mediterranean diet, which this audit backs as reproducibly as anything in the longevity field; exercise; and sleep — the MULTI Consortium found a U-shaped relationship between sleep and biological ageing across 23 clocks, with an optimum around 6.4 to 7.8 hours a night. Not the eight-plus hours of wellness dogma. Not the five of ambition. Seven-ish.

None of that requires a cheek swab. You do not need a clock to tell you what to do. You need a clock honest enough to tell you whether what you're doing works — and for the first time, we have one.

Your biological age is a number. After this week, it might finally be the right one.

Sources

  1. Sehgal, R., Borrus, D., Armstrong, J.F., et al. "Responsiveness of epigenetic aging biomarkers to longevity interventions in humans." Nature Medicine (2026). DOI: 10.1038/s41591-026-04562-9 (21 Aug 2026).
  2. Horvath, S. "Putting epigenetic aging clocks on trial." Nature Medicine, News & Views (2026). DOI: 10.1038/s41591-026-04524-1 (21 Aug 2026).
  3. Fleming, T.R. & Powers, J.H. "Biomarkers and surrogate endpoints in clinical trials." Statistics in Medicine 31, 2973–2984 (2012). DOI: 10.1002/sim.5403.
  4. Belsky, D.W. et al. "DunedinPACE, a DNA methylation biomarker of the pace of aging." eLife 11, e73420 (2022). DOI: 10.7554/eLife.73420.
  5. MULTI Consortium et al. "Sleep chart of biological ageing clocks in middle and late life." Nature (2026). DOI: 10.1038/s41586-026-10524-5.