A new Nature Medicine study put lifestyle programs, prescription drugs, supplements, and procedures through the same set of biological-age tests. The results are more useful than exciting, and that's a good thing.
"Biological age" has become one of the most marketed phrases in health.
You can buy a test that tells you your cells are 41 when your birth certificate says 52. You can buy a supplement stack that promises to bring that number down. And you can find a clinic that will offer an infusion, a chamber, or a protocol to do the same.
What has been missing is a fair comparison. Most of these claims come from single studies, each using a different test, measured in different people over different time frames. It's hard to compare a diet trial that used one clock with a drug trial that used another. Until recently, the honest answer to "does any of this work?" was that we didn't have the data to say.
A new paper in Nature Medicine doesn't settle the question. It does give us the first reasonable way to ask it.
What they did
Raghav Sehgal, Albert Higgins-Chen, and colleagues at Yale built a harmonized database called TranslAGE. It contains 51 longitudinal human intervention studies, both public and private, each with blood DNA samples taken before and after an intervention (PubMed).
For every study, they calculated the same 16 prominent epigenetic clocks plus 94 additional DNA methylation biomarkers. They adjusted for chronological age, standardized the results for comparison across studies, and then examined changes between the before-and-after samples (PubMed).
A brief explanation of what's being measured. Epigenetic clocks look at DNA methylation, chemical tags that attach to DNA and help control how genes behave without changing the genetic code itself. These patterns shift predictably as we age, and algorithms can use them to estimate a "biological" age. Some newer clocks reflect specific processes such as inflammation, immune function, or the pace of physical decline (SciTechDaily/Yale).
The interventions fell into four groups: lifestyle changes, prescription drugs, over-the-counter supplements, and medical procedures.
What moved the clock
Exercise plus a healthy diet. Programs that combined exercise with a healthy eating plan consistently lowered epigenetic age. And it didn't matter much whether that plan was Mediterranean, low-carb, or low-fat (SciTechDaily/Yale). Diet interventions also showed up in the DNA-based proxies for metabolic markers. The epigenetic signal for triglycerides was significant in six of seven dietary interventions, and the glucose proxy in four of seven (News-Medical).
That finding deserves more attention than it will get. The internet has spent years arguing about which diet is the anti-aging diet. In this dataset, the specific diet was less important than having one and pairing it with exercise.Certain prescription drugs. Pharmacological interventions had the largest average effects, larger than lifestyle (News-Medical). The strongest results came from three places:
Metformin, the long-standing diabetes drug, where the more detailed clocks showed especially large changes in inflammatory, brain, and metabolic system scores (News-Medical).
Semaglutide, the GLP-1 medication used for diabetes and weight management (SciTechDaily/Yale).
Anti-TNF therapy, drugs that block tumor necrosis factor, an immune protein that drives inflammation. In people with arthritis or inflammatory bowel disease, these drugs shifted almost all of the second-generation clocks to a similar degree (News-Medical).
The common thread is metabolism and inflammation. The authors point to pathways involving AMPK, mTOR, and TNF as possible explanations, though they present these as hypotheses rather than established mechanisms (News-Medical).
What didn't
The over-the-counter supplements included in the analysis, and some medical procedures, did not significantly reduce epigenetic age (SciTechDaily/Yale). Senolytics, compounds meant to clear aging "zombie" cells, gave inconsistent results across five studies (News-Medical).
Overall, 19 interventions significantly decreased epigenetic aging measures, five significantly increased them, and most of the rest showed no significant effect (News-Medical).
I want to be fair here. "The supplements studied" is not the same as "all supplements," and a null result in a small study isn't proof that something does nothing. But the gap between what these products promise and what they showed on the same measurement that works for exercise and metformin is worth keeping in mind.
Two findings that change how to read this
Sick people changed more than healthy people. Biomarkers moved more in participants with existing disease than in healthy volunteers (SciTechDaily/Yale). The likely explanation is that there's more age-related disruption to improve. That matters for interpretation. A drug that lowers the clock in someone with uncontrolled diabetes or active Crohn's disease may simply be treating the disease well. That's valuable, but it isn't the same as slowing aging in a healthy 45-year-old.
The choice of clock matters. The clocks trained to predict mortality or pace of aging, including DunedinPACE, PCGrimAge, GrimAgeV2, SystemsAge, and PCPhenoAge, were the most responsive and agreed with each other the most. Older, first-generation clocks, built mainly to predict chronological age, did worse (PubMed; News-Medical). If you've had a consumer biological-age test, it's reasonable to ask which clock it used.
Why this matters
Proving that something extends a healthy human life takes decades. That is the central problem in longevity research, and it's why so much of the field runs on mouse data and marketing.
If a biomarker could reliably stand in for long-term outcomes, we could test interventions in months or a few years. This paper is groundwork for that. It shows which biomarkers respond to which interventions, in which people, so future trials can choose their endpoints deliberately instead of hoping for the best (PubMed).
That is a real contribution. It is also a first step.
What this doesn't say
The senior author, Dr. Higgins-Chen, was direct about this. Moving a biomarker in a younger direction does not automatically prove that an intervention slows aging, prevents disease, or extends life (SciTechDaily/Yale). Several questions remain open:
Are the changes caused by the intervention, or by confounding factors?
When do the effects start, and do they last after treatment stops?
What biological mechanisms produce them?
Most importantly, does a younger clock actually mean lower risk of disease and functional decline?
A few other points worth noting. This is a pooled analysis of existing studies, many of them small and designed for different purposes. The researchers want larger and more diverse populations next, because age, health status, genetics, and medications may all affect how biomarkers respond (SciTechDaily/Yale). The authors also disclose commercial ties to this field. Two are co-inventors on a patent for one of the clocks, some have consulted for epigenetic testing companies, and three co-authors are employees of TruDiagnostic, which contributed data (PubMed). That doesn't invalidate the work, and it's common in a young field. It is context you should have.
What I'd take from it
I would not start metformin or semaglutide to lower a biological-age number. Both are real medications with real side effects and real indications, and this study doesn't establish that they slow aging in healthy people. If you have a condition they treat, that's a conversation with your clinician on its own merits.
I also wouldn't spend much money on biological-age testing yet, except out of curiosity. The number may move, but we don't yet know what that movement means for you.
What I would take from it:
Exercise and a sensible diet remain the foundation. The evidence supports them across diet types, so you can choose the pattern you'll actually follow.
Treating metabolic and inflammatory disease well is longevity medicine. Blood sugar, weight, and chronic inflammation showed the strongest signals
.Be skeptical of supplements sold as anti-aging. The ones tested here didn't measure up.
Watch this space. If these clocks are validated against real outcomes, they could change how quickly we learn what works.
The answer isn't new. What's new is that we can finally measure the claims against each other on the same terms, and the familiar answers held up.
Sehgal R, Borrus D, Armstrong JF, et al. Responsiveness of epigenetic aging biomarkers to longevity interventions in humans. Nature Medicine. 2026. doi:10.1038/s41591-026-04562-9. PubMed
Plain-language summary: Yale University via SciTechDaily
