Diet and Exercise Moved 16 Aging Clocks. Supplements Didn’t

Yale researchers took 51 studies that had tried to slow human aging, from Mediterranean diets to metformin to fish oil to hyperbaric oxygen, and ran every one of them through the same 16 epigenetic clocks. In total, 3,128 blood samples. The question was simple and overdue: when you measure everything the same way, what actually moves the needle? The answer, published in Nature Medicine on August 21, 2026, is good news for the gym and the grocery aisle and bad news for the supplement shelf. It is also more complicated than the slogan it has been turned into.

One ruler for 51 experiments

Until now, every aging study picked its own clocks, often one or two, and reported whichever moved. That makes comparison nearly impossible and false positives easy. The team behind the paper, led by Raghav Sehgal and Albert Higgins-Chen, built a harmonized database called TranslAGE from public repositories and from private datasets held by TruDiagnostic, a company that sells epigenetic tests. For each study they computed the same 16 prominent clocks plus 94 other methylation markers, removed the effect of calendar age, standardized the results so studies could sit side by side, and compared blood drawn before and after each intervention.

They sorted interventions into four groups: lifestyle programs, built mostly around diet and exercise; prescription drugs such as metformin, semaglutide and TNF blockers; supplements such as omega 3 fatty acids and folate; and medical procedures, from hyperbaric oxygen to organ transplants. Forty one of the 51 studies were in healthy people, ten in people with a disease.

What moved, and how much

Only two categories lowered epigenetic age across the 16 clocks: drugs and lifestyle. Supplements and procedures did not. Drugs had the larger average effect, about 0.09 standard deviations, against roughly 0.04 for lifestyle. Taken one by one, 19 interventions lowered clock readings overall, 13 of them after correcting for multiple comparisons, five raised them, and 26 showed no significant effect.

The authors give a way to translate those units. For widely used clocks, they write, an effect of about 0.2 standard deviations corresponds to roughly one year of epigenetic age. By that yardstick, which is my rough arithmetic and not theirs, the average lifestyle effect works out to something like two or three months. Real, measurable, and nowhere near the "ten years younger" claims sold online.

Consistency is where diet stands out. The team set two rules for a believable effect: clocks of the same generation should move together in one study, and a second study of the same intervention should move the same clocks. TNF blockers passed both. So did two different Mediterranean diets tested in healthy people. Five senolytic studies failed, with clocks pointing in opposite directions within and across studies. Across seven diets, from low carb to vegan, one organ system score, musculoskeletal, dropped in every single one.

Drugs moved the clocks furthest. Diet moved them most reliably. Supplements barely moved them at all.

The clocks themselves were graded too. Newer clocks trained to predict mortality or the pace of aging responded best. DunedinPACE, which estimates how fast a body is aging rather than how old it is, fell significantly in 16 interventions and rose in one, and it was the most responsive clock to lifestyle programs, dropping in 8 of 15.

The fine print on "basics win"

Three caveats keep this from being a clean victory lap. First, drugs beat lifestyle on size, and the authors say so. The headline that exercise and diet "beat" everything only holds against supplements and procedures. Second, the strongest responses came from sick people, who have more room to improve. In the healthy studies, a few clocks dipped, but none survived correction for multiple comparisons. Third, a clock reading is a surrogate. The paper notes there is still no agreed minimal change in these clocks that means anything clinically, and that clocks trained partly on inflammation markers may simply be tracking inflammation.

The authors also call their pooling descriptive, not causal: the 51 studies differ in age, duration, design and quality, and the samples were not batch corrected together. Several authors have commercial ties. Two are named inventors on a patented clock in the panel, two have consulted for TruDiagnostic, and three are TruDiagnostic employees who developed another of the clocks tested.

This matters for the companion claim of the season: that the basics beat a $2 million protocol. Bryan Johnson reportedly spends about $2 million a year on his regimen. In 2023, Fortune reported that a 55 year old recruiter spending $27 a month on a gym and $79 on a supplement subscription ranked No. 2 on the Rejuvenation Olympics, a DunedinPACE leaderboard Johnson cocreated, while he sat at No. 6. That story has been recirculating this fall as fresh news. It is three years old, her budget included supplements from a company whose study she joined, and when I checked the leaderboard on October 13, 2026, Johnson was No. 1 at 0.50. The leaderboard is run in partnership with TruDiagnostic, which sells the tests at $249.50 each on an annual plan.

No study has compared the Johnson protocol with an ordinary healthy routine, and no leaderboard of self selected testers can. What the Yale data can say is narrower and sturdier: among things anyone can do, diet and exercise are the only category that moved these clocks consistently, and the pills sold over the counter did not. For how to judge single clock claims, see my breakdown of why biological age claims need more than one clock, which comes from the same Yale lab.

What comes next

Sehgal told Yale News the next step is testing in a larger and more diverse group of volunteers, and the paper says future versions of TranslAGE will add batch adjustments where data allow. The real milestone is further out: tying a given drop in DunedinPACE or GrimAge to fewer heart attacks, less frailty or longer life. Until then, every clock result, including the good ones for diet, is a promising surrogate rather than a verdict.

What strikes me most is how small and how cheap the reliable signal is. The thing that worked across studies is the thing your grandmother would have prescribed, and it bought a few months on a lab readout, not a decade. That is not a disappointment. It is a calibration. Anyone promising you far more than a Mediterranean diet and regular exercise can deliver on these clocks now has to explain why their product would beat a category that already beats supplements, using the same ruler, in front of everyone.