I’ve been side-eyeing biological-age tests for years, and when I opened this paper I still couldn’t answer a basic question: if a pill slowed aging, how would anyone know before the funeral?

There is no direct evidence linking change in an epigenetic clock to future outcomes.
David Furman, Buck Institute

ScienceDaily’s answer, under the headline “Yale scientists reveal what really slows aging and what doesn’t,” is tidy. Certain prescription drugs, healthy diets and exercise measurably slow biological aging, it says, and popular anti-aging supplements appear largely ineffective. The paper underneath supports something narrower. When the authors asked which treatments moved the same biomarkers again in a second study of the same intervention, only two passed their consistency test: TNF-blocking drugs in people with inflammatory arthritis or inflammatory bowel disease, and two different Mediterranean diets in healthy people.

Neither result shows anyone living longer. Both show a blood test that reads chemical tags on DNA, and the tags shifting the same way twice. The headline skips that distinction, and you’d want it before spending a dollar on any of this.

The work, led by Raghav Sehgal with senior author Albert Higgins-Chen of the Yale School of Medicine, appeared in Nature Medicine in August, after a preprint posted in 2024. The team didn’t draw new blood. Instead it pooled existing before-and-after DNA methylation data from 51 longitudinal intervention studies in humans (diets, exercise programs, drugs, supplements and medical procedures) into one harmonized database called TranslAGE, then scored every study against the same 111 DNA methylation biomarkers, 16 of them epigenetic clocks. “Our field has developed an extraordinary number of biological aging biomarkers,” Sehgal said. This is the first serious attempt I’ve seen to grade them against each other on the same data.

BY THE NUMBERS
51
human intervention studies
111
DNA methylation biomarkers
16
epigenetic clocks
The scale of the pooled TranslAGE analysis. Source: Yale School of Medicine: Yale researchers create roadmap to measure whether longevity interventions work (2026)

Across the pool, 19 interventions significantly lowered DNA-methylation aging measures and 5 significantly raised them, while most of the rest did nothing detectable. Lifestyle and drug interventions produced the strongest drops, and drugs posted the larger average effects. ScienceDaily lists metformin and semaglutide among the treatments tied to the strongest reductions. The five senolytic studies, the “zombie cell” clearing compounds the longevity crowd loves, pointed in divergent directions.

SIGNIFICANT SHIFTS IN DNA-METHYLATION AGING MEASURES (interventions)
Figure data
GroupValue (interventions)
Lowered19
Raised5
Most of the remaining interventions showed no significant effect. Source: News-Medical: Scientists compared 51 longevity studies to see which aging biomarkers respond most reliably (2026)

Put out the fire and the blood looks younger

So why would a drug for swollen joints make a methylation clock read younger? A clock doesn’t know what an injection is. It reads tags on DNA in white blood cells, and the biomarkers that responded most consistently were the ones designed to measure the pace of aging or predict mortality, not birthdays. TNF is one of the body’s loudest inflammatory signals. TNF-targeting therapies shifted almost all of the second-generation biomarkers to a similar extent in people with arthritis or bowel disease, and several biomarkers dropped further in studies of people with disease than in healthy populations. The authors suggest that the population each biomarker was trained on partly explains who responds: clocks built on sicker, more varied groups pick up changes in sick people. My own reading, and it’s mine rather than theirs, is this: calm a fire in someone’s joints and their blood stops looking like it’s burning. Whether that person is aging more slowly is a separate question, and this analysis wasn’t built to answer it.

Which makes the Mediterranean result stranger, and better. Two diets lowered similar second-generation biomarkers in healthy individuals, people with no inflammatory blaze for a drug to douse. That is the only repeat result in the pool that came from people who weren’t already sick, and it came off a plate.

“Drugs win” makes a much better headline for one industry. Metformin is a decades-old generic. Semaglutide is Novo Nordisk’s blockbuster, and an anti-aging halo would be a gift to its marketing. The best-known clock result for the drug is a post hoc analysis of 84 people with HIV-associated lipohypertrophy, drawn from a trial that randomized 108, followed for 32 weeks. Against placebo, semaglutide was associated with a 3.08-year smaller annual increase in PCGrimAge and a DunedinPACE pace about 9 percent slower. The authors call the epigenetic outcomes exploratory and did not adjust p-values for multiple comparisons, and two of the seven authors are TruDiagnostic employees. It’s a signal worth chasing in a small, unusual population. It’s no reason for a healthy 45-year-old to start a weekly injection.

A testing lab helped supply the evidence

That company keeps turning up. TruDiagnostic is an epigenetic testing lab, and three of the Yale paper’s co-authors, Ryan Smith, Natalia Carreras and Varun B. Dwaraka, are listed as TruDiagnostic collaborators who contributed data and/or algorithms. The TranslAGE database draws on TruDiagnostic clinical trials. The disclosed funding comes from the National Institute on Aging (R01AG065403), an Impetus Grant and two Yale fellowships, but grant money and company data are separate questions, and a lab in the clock business helped supply the evidence about which clocks respond. Keep that in mind the next time an ad quotes this paper at you.

Higgins-Chen frames the project as a guide to measurement: “choosing the right biomarker becomes just as important as choosing the right intervention.” Fair enough. But a paper about choosing measuring tools got sold as a paper about what slows aging.

The supplement verdict outran the data too. Sehgal himself told NutraIngredients: “The appropriate conclusion is not that supplements do nothing, nor that a particular supplement is an anti-aging treatment.” Some of the strongest supplement trials weren’t even in the pool. DO-HEALTH and COSMOS both reported statistically significant but modest effects on newer clocks, and neither made it into TranslAGE because of data availability limits. DO-HEALTH’s post hoc analysis of 777 older adults tested 2,000 IU of vitamin D, 1 gram of omega-3 and a home exercise program over three years, and omega-3 alone slowed PhenoAge, GrimAge2 and DunedinPACE, an effect its authors called small. The smaller, shorter nutrient-combination studies that did make the pool gave mixed results, in one case moving the same biomarkers in inconsistent directions. That’s inconclusive. It isn’t a loss, and it isn’t a win either.

David Furman of the Buck Institute gave the fairest summary of the category: “Individual nutraceuticals rarely move composite biological age markers reliably at the population level.” Then he added the line I’d print on every biological-age test kit: “There is no direct evidence linking change in an epigenetic clock to future outcomes.”

That warning covers metformin, semaglutide, fish oil and olive oil alike. ScienceDaily’s own write-up concedes that the biomarker changes “do not necessarily mean people will live longer or avoid age-related diseases,” that nobody yet knows whether the treatments caused them, and that nobody knows whether they persist after treatment ends.

So the list of things that moved the same markers twice is short: a drug class that calms inflammation in people who have plenty of it, and a way of eating that held up in healthy people. Nobody holds a patent on lentils and olive oil.

I won’t pay a lab to grade my supplements, and I won’t take a prescription drug to nudge a methylation number. I’ll keep cooking the Mediterranean way, because it’s the only thing in these 51 studies that lowered similar markers twice in healthy people, and I’m not waiting for a clock to tell me it’s working.

Sources

  1. Nature Medicine: Sehgal, Higgins-Chen et al., “Responsiveness of epigenetic aging biomarkers to longevity interventions in humans” (2026)
  2. Yale School of Medicine: Yale researchers create roadmap to measure whether longevity interventions work (2026)
  3. ScienceDaily: Yale scientists reveal what really slows aging and what doesn’t (2026)
  4. News-Medical: Scientists compared 51 longevity studies to see which aging biomarkers respond most reliably (2026)
  5. Lifespan.io: Comparing the Responsiveness of Epigenetic Aging Biomarkers (2026)
  6. NutraIngredients: Epigenetic biomarkers show promise for longevity, but which clock moves which needle? (2026)
  7. Corley, McComsey et al., Semaglutide and epigenetic aging in people with HIV-associated lipohypertrophy, randomized-trial analysis (2025)
  8. Nature Aging: DO-HEALTH trial, omega-3, vitamin D and exercise effects on epigenetic aging clocks (2025)
  9. bioRxiv: DNAm aging biomarkers are responsive: Insights from 51 longevity interventional studies in humans (preprint, 2024)