Evidence guide · South Africa

Do anti-ageing supplements lower biological age? What a 51-study epigenetic clock analysis found

A Nature Medicine analysis of 51 intervention studies found drugs and lifestyle changes lowered epigenetic clock readings while the supplements in the database did not. Which supplements were in it, what the drug results mean, and why one clock reading proves little.

Review type: research summary. This page summarises published research and product labels, cited below.

Epigenetic clocks estimate "biological age" from chemical tags (methyl groups) on DNA in a blood sample, and they have gained traction as outcome measures for anti-ageing interventions. A Yale-led analysis published in Nature Medicine on 21 August 2026 pooled 51 human intervention studies and recalculated the same panel of clocks for each one. Drug and lifestyle interventions, as groups, lowered epigenetic age. The supplements in the database, as a group, did not.

Short answer: this is not a verdict on every anti-ageing supplement. The supplement studies in the database were mostly small and short, and they did not include NMN, nicotinamide riboside, urolithin A or spermidine, so the analysis says nothing about those products either way. It also cannot show that a lower clock reading means a longer or healthier life. What it does show is that clock results vary a lot by clock, by study and by who was studied, which is a good reason to treat a single "you are younger" test result with caution.

What the Yale analysis did

The team built a database called TranslAGE from 51 longitudinal intervention studies with blood DNA methylation measured before and after treatment, a total of 3,128 samples. The data came from public repositories (GEO and EMBL) and from private datasets held by TruDiagnostic, a company in Lexington, Kentucky, where three of the authors work. For every study the authors calculated the same 16 widely used epigenetic clocks plus 94 other DNA methylation biomarkers, adjusted for chronological age, standardised the results so studies could be compared, and compared readings before and after treatment. They grouped interventions into four categories: lifestyle (diets and exercise), pharmacological (such as metformin, rapamycin, semaglutide, ketamine and anti-TNF drugs), supplements, and medical procedures (such as hyperbaric oxygen, organ transplants and gene therapy).

What it found

  • By category, only pharmacological interventions (mean effect size -0.093) and lifestyle interventions (-0.039) significantly reduced epigenetic age. Pharmacological interventions had larger effects than supplements (P = 0.0009) and than lifestyle interventions (P = 0.0063).
  • Across all 51 interventions, 19 significantly decreased epigenetic age across the 16 clocks (13 after correction for multiple testing), 5 significantly increased it (3 after correction) and 26 had no significant effect.
  • The most consistent results came from anti-TNF drugs in people with arthritis or inflammatory bowel disease, which shifted nearly all second-generation clocks in the same direction across studies, and from two Mediterranean-style diet programmes which the authors describe as healthy cohorts. Those diets decreased similar sets of second-generation clocks across the two studies. In the paper's supplementary dataset table, both programmes also included exercise, and the participants had abdominal adiposity or dyslipidaemia.
  • Several DNA methylation biomarkers showed larger decreases in studies of people with a diagnosed disease than in people without one. DunedinPACE responded similarly in both groups. The authors say the larger decreases probably reflect higher baseline biological dysregulation, and that this fits earlier findings of more room for improvement where the body is more impaired.
  • Newer clocks such as DunedinPACE, PCGrimAge and GrimAgeV2, built to track mortality or the pace of ageing, and the explainable clock SystemsAge, responded most consistently. First-generation clocks (Horvath1, Horvath2 and Hannum) rose or fell sporadically without a clear pattern. The authors write that their data support earlier reports that generation 1 clocks such as Horvath1 largely reflect stochastic epigenetic noise.

The supplements in the database

The paper's supplementary dataset table lists 11 supplement datasets. They covered 800 micrograms of folic acid a day for eight weeks; vitamin B12, folic acid and vitamin D3 taken together for two years; 3 g a day of omega-3 fats for six weeks in people with obesity; 200 mg a day of oligomeric proanthocyanidins (OPCs); 700 mg a day of cat's claw (Uncaria tomentosa) extract; a buckwheat extract; a freeze-dried human milk product; a lavender oil product; an unnamed botanical supplement; and a multi-capsule supplement protocol measured at two time points. Each dataset had between 10 and 50 participants.

In the per-intervention results, three of these datasets (cat's claw extract, the human milk product and the unnamed botanical) showed an overall decrease across the 16 clocks, the buckwheat extract showed an overall increase, and the rest showed no significant change. The authors flag the human milk product as one where only a single reliable newer clock moved, a pattern they say may point to a false positive. For the two time points of the multi-capsule protocol, the same clocks changed but in inconsistent directions.

None of the supplement datasets involved NMN, nicotinamide riboside, urolithin A, spermidine, resveratrol or creatine. Yale's news release says the over-the-counter supplements in this analysis had little effect on epigenetic age. Reading that as a claim that supplements do not slow ageing goes further than the data: the supplement set was small and mixed, and most of the studies were brief, with 10 to 50 participants.

The drugs: a closer look

In their discussion, the authors single out anti-TNF therapies and metformin as producing consistent, strong changes. Both were studied in people with a disease: the anti-TNF datasets came from people with rheumatoid arthritis or inflammatory bowel disease, and the metformin dataset was six people with HIV. Yale's news release lists semaglutide alongside metformin as among the interventions with the largest effects. In the paper's per-intervention table, the semaglutide dataset, 11 people treated for about six months, did not show a significant overall change across the 16 clocks.

Of the two weekly rapamycin datasets, one (nine people) showed no significant overall change and the other (seven people) showed an overall increase in epigenetic age. Across five senolytic datasets using dasatinib and quercetin, with or without fisetin, clocks moved in different directions, and the authors write that senolytic effects "may be inconsistent" from an epigenetic ageing perspective.

What it cannot show

  • The authors write that it is still unclear whether short-term biomarker changes correspond to long-term improvements in disease, healthspan or lifespan.
  • The authors write that there is no empirically defined minimal clinically important difference for these clocks, and that this is a key barrier to using them as surrogate endpoints. For clocks such as Horvath, Hannum, PhenoAge and GrimAge, they estimate that an effect size of about 0.2 corresponds to about a one-year shift in epigenetic age, given a standard deviation of about 4 to 6 years.
  • The pooled studies differed widely in age range, treatment, duration and design, and their quality and size varied. The authors describe their aim as descriptive, not a test of whether individual treatments work.
  • Second-generation clocks need cautious reading because they were trained on composite risk factors, for example C-reactive protein in GrimAgeV2, so a change may reflect a shift in inflammation or another health marker rather than a direct effect on ageing biology.
  • Two authors are co-inventors of SystemsAge, one of the clocks in the panel, which is patented in the US. The senior author has received consulting fees from TruDiagnostic and FOXO Biosciences, and the first author from TruDiagnostic, LongevityTech.fund and Cambrian BioPharma. Three authors are TruDiagnostic employees who developed OMICmAge, another of the clocks analysed. The work was supported by the US National Institute on Aging, with other grants and fellowships also listed.

Single-clock claims deserve scepticism

A companion paper from the same Yale lab, published in Aging Cell in October 2026, re-analysed earlier intervention studies that had reported a result on a single clock. When additional clocks were added, none of them reached statistical significance. The authors propose three tests before trusting a result: reliable principal-component-based clocks respond, clocks trained on mortality or pace of ageing respond, and several clocks move in the same direction. A product or clinic that quotes one favourable clock reading, especially from a first-generation clock, does not meet that standard.

Larger trials tell a more modest story

Two larger randomised trials sit outside the Yale database. In a post hoc analysis of the DO-HEALTH trial, 777 older adults took 2,000 IU of vitamin D a day, 1 g of omega-3 a day, a home exercise programme, or combinations, for three years. Omega-3 alone slowed PhenoAge, GrimAge2 and DunedinPACE, and the three treatments had additive benefits on PhenoAge. The standardised effects were 0.16 to 0.32 units, about 2.9 to 3.8 months. One author, Steve Horvath, is a founder of the nonprofit Epigenetic Clock Development Foundation, which has licensed several patents for epigenetic clocks, including GrimAge, from UC Regents, and he works for Altos Labs.

In the CALERIE trial, 220 adults without obesity were randomised to 25% calorie restriction or an unrestricted (ad libitum) control diet for two years. Calorie restriction slowed the pace of ageing measured by DunedinPACE but did not significantly change biological age estimates from clocks including PhenoAge and GrimAge, and the effect sizes were small. The contrast with the six-week omega-3 dataset in the Yale database fits the Yale finding that study duration and population shape whether clocks move.

Our view

  • The Yale paper is mainly a test of the clocks, not of supplements. Its clearest message is that newer clocks such as DunedinPACE and PCGrimAge respond more reliably than older ones, and that one-clock results are weak evidence.
  • The supplement category in the database was small and narrow. It does not show that popular longevity supplements such as NMN or urolithin A work, and it does not show that they fail. They were not in it.
  • The drug signals come mostly from people treated for a disease. They are not a reason for healthy people to take metformin, semaglutide or anti-TNF drugs for ageing, and the paper does not suggest that.
  • In this research, the lifestyle results with the most consistent clock changes are the Mediterranean-style diet programmes, which also included exercise, and, from separate trials, a small DunedinPACE change with calorie restriction and a small change on three clocks with omega-3 over three years. These are biomarker shifts, and a link to longer life is not yet shown.
  • If you buy a biological age test, ask which clock it uses, treat a single reading cautiously, and do not change medicines because of it.

South African rules

In South Africa, health supplements are regulated by SAHPRA as complementary medicines under the Medicines and Related Substances Act 101 of 1965, and SAHPRA publishes guidance on the safety and efficacy of health supplements. SAHPRA's complementary-medicines page limits health supplements to low-risk indications, such as general health maintenance, without reference to a specific disease. A label claim that a supplement reverses biological age is not supported by the evidence above.

FAQ

Do NMN or NAD+ supplements lower biological age?

This analysis cannot say. No NMN, nicotinamide riboside or other NAD+ precursor study was among its supplement datasets. Claims that these products reverse epigenetic age need their own trials, measured on reliable newer clocks and repeated across studies.

Is a home biological age test worth it?

It can be interesting, but it is not a diagnosis. The Yale work found that clocks do not always agree, that generation 1 clocks such as Horvath1 largely track stochastic epigenetic noise, and that there is no empirically defined threshold for a meaningful change. If you test, find out which clock is used, and compare like with like over time rather than reading meaning into one result.

Does semaglutide slow ageing?

That has not been shown. Yale's news release lists it among the drugs with the largest effects, but in the paper's per-intervention table the semaglutide dataset, 11 people, did not show a significant overall change across the 16 clocks, and no clock change has been shown to translate into longer life.

Should I take omega-3 for ageing?

DO-HEALTH found a small slowing of three clocks over three years with 1 g a day in older adults, in a post hoc analysis. That is a biomarker result, not proof of longer or healthier life. Ask your doctor or pharmacist before starting a supplement, especially if you take other medicines.

Sources

These are the documents this page relies on.

  1. Responsiveness of epigenetic aging biomarkers to longevity interventions in humans. Nat Med. 2026;32(9):3477-3490 (online 21 August 2026)Nature Medicine (PubMed 42629466)
  2. Supplementary Table 1 (TranslAGE intervention names, categories, regimens, subjects and length) for Responsiveness of epigenetic aging biomarkers to longevity interventions in humansNature Medicine
  3. Supplementary Table 5 (one-sample t-test per intervention across 16 clocks) for Responsiveness of epigenetic aging biomarkers to longevity interventions in humansNature Medicine
  4. Can you slow the aging process? Study reveals which interventions might helpYale News
  5. When to Trust Epigenetic Clocks: Avoiding False Positives in Aging Interventions. Aging Cell. 2026;25(10):e70756Aging Cell (PubMed 42850743)
  6. Individual and additive effects of vitamin D, omega-3 and exercise on DNA methylation clocks of biological aging in older adults from the DO-HEALTH trial. Nat Aging. 2025;5(3):376-385Nature Aging (PubMed 39900648)
  7. Effect of long-term caloric restriction on DNA methylation measures of biological aging in healthy adults from the CALERIE trial. Nat Aging. 2023;3(3):248-257Nature Aging (PubMed 37118425)
  8. Complementary medicinesSAHPRA
  9. Complementary medicines: health supplements safety and efficacySAHPRA