# The best biological age clocks predicted death in 18,859 people URL: https://sciencebased.news/biological-age-tests/ Published: 2026-10-12 Updated: 2026-10-19 Claim: The newest biological age clocks predict who will get sick and who will die sooner. Grade: Promising What to do: - If you test, choose a lab that reports GrimAge or DunedinPACE. Over 10 years in 18,859 adults, the newest clock was linked to 37 of 174 diseases, the first clocks to 1 each. - Use blood, not saliva or a cheek swab, at the same time of day each time. Judge three or more tests, not one. The time of day alone moved one man's result by about 3 years. - Keep tracking the numbers with the hardest proof too, such as blood pressure, cholesterol, blood sugar and fitness. - Safety: a high result is not a diagnosis and a low one is not a clean bill of health. No home biological age test is FDA-approved as a medical test. - Not proven: that lowering your score makes you live longer. Two trials moved the clocks a little, and neither one counted deaths. # The best biological age clocks predicted death in 18,859 people *Evidence check, October 12, 2026. Updated October 19, 2026.* > **The claim:** The newest biological age clocks predict who will get sick and who will die sooner. > > **Our grade: Promising.** In 18,859 adults followed for 10 years, the newest clock was linked to 37 of 174 diseases and to a 54% higher risk of death for each step it ran ahead. It is one large observational study, a single home reading can still swing by years, and no trial shows that a lower score means a longer life. ![A woman in her 30s blowing out the candles on a birthday cake, with a blood collection tube in a rack beside the cake](https://sciencebased.news/images/biological-age-tests-hero-2.jpg) Maya is 38. A small box on her kitchen table holds a home test that promises to tell her how old her body really is. Can a number like that mean anything? It can. When researchers ran 14 of these clocks on blood from 18,859 Scottish adults and waited 10 years, the newer clocks predicted who died and who got sick [1]. ## Where the claim came from Your DNA carries chemical tags, called methylation, that help switch genes on and off. The pattern shifts as you age. An "epigenetic clock" is a formula that reads those tags from blood or saliva and turns them into an age, or into a speed of aging. The first clocks, from 2013, were built to guess your birthday age. Later ones were built on health instead: GrimAge on who died, DunedinPACE on how fast people's organs declined as they got older [7]. Bryan Johnson made the tests famous. In February 2025, writing about his own company's test, he said DunedinPACE "reports the speed of aging which makes it very responsive to interventions and ideal for evaluating them" [9]. He tops a public leaderboard, the Rejuvenation Olympics, with an average pace of 0.503 when we checked on October 12, 2026 [11]. On that scale 1.0 means one year of aging per calendar year. Peter Attia gave the clocks a full episode on April 6, 2026, and called them exciting tools [8]. ## From guessing birthdays to predicting disease The Scottish study is the fairest test so far, because it ran every clock on the same people. ![Bar chart of how many of 174 diseases each aging clock was linked to over 10 years in 18,859 Scottish adults, after accounting for age, sex and lifestyle. Hannum and Horvath, built to guess age: 1 each. PhenoAge 22, GrimAge 29, DunedinPACE 33, GrimAge version 2: 37.](https://sciencebased.news/images/biological-age-tests-chart.png) The team tracked 174 diseases. After accounting for age, sex, weight, smoking, drinking, education and neighborhood poverty, the two oldest clocks, Hannum and Horvath, were each linked to 1 disease. The clocks built on health did far better: PhenoAge was linked to 22, GrimAge to 29 and DunedinPACE to 33. GrimAge version 2, the newest, was linked to 37 [1]. It also did best on death. Over the 10 years, 842 people died. For each standard step that a person's clock ran ahead of their real age, the risk of death was 54% higher [1]. Across the newer clocks the strongest links were with lung and liver conditions [1]. How much does the clock add to what a doctor already knows? A model using only those ordinary facts sorted who would die with a score of 0.851 out of 1. Adding the best clock raised it to 0.865 [1]. That is a small step, but it comes on top of your age, your weight and whether you smoke. The clock saw something those facts did not. One of the authors invented DunedinPACE, which is licensed to the testing company TruDiagnostic. The study was publicly funded [1]. ## The newer clocks also repeat better A test is only useful if it gives the same answer twice. When a lab ran the same blood samples twice, older clocks disagreed with themselves by up to 9 years. Rebuilt versions of the clocks agreed within about 1.5 years on most repeat runs [2]. In a 2026 analysis, DunedinPACE and the rebuilt GrimAge were among the steadiest, though two of its authors have been paid consultants to TruDiagnostic [5]. ## The score can move Two randomized trials have moved it a little. In CALERIE, 197 healthy adults either ate as usual or aimed to cut calories by 25% for two years. Dieting slowed DunedinPACE by 2% to 3%. Two other clocks, PhenoAge and GrimAge, did not move, and the results of the two groups overlapped by close to 90% [6]. In DO-HEALTH, 777 Swiss adults over 70 were followed for three years. One gram of omega-3 a day slowed three of four clocks. On two of them the gain was about 3 to 4 months of aging; on DunedinPACE it was about 1%. The clock analysis was added after the trial was done, which makes it weaker evidence [7]. Neither trial counted deaths, so no one has yet shown that a lower score means a longer life. ## How to get a reading you can trust A result that holds across 18,859 people can still be noisy for one of them. Three things move a reading that have nothing to do with how you are aging, and each has a fix. The clock is one. Older clocks read the same blood sample up to 9 years apart, so pick a lab that reports the newer ones [2]. The hour is another: one man's reading rose about 3 years from late night to late morning, so test at the same time of day [3]. The third is what you send in. Saliva and cheek swabs read far older than blood from the same person, by almost 30 years on some clocks, so use blood [4]. The lab study ran 36 blood samples twice each. The well-known Horvath clock differed by a median of 1.8 years between the two runs, and by 4.8 years at most. Across six clocks the worst gaps were 3 to 9 years. Two of them have clocks licensed to the test company Elysium Health, which supplied some samples [2]. The time-of-day study drew blood from one healthy 52-year-old man every 3 hours for 72 hours. On the Horvath clock he read 55.3 at about 11:30 at night and 58.5 at about 11:30 in the morning. Of 17 clocks, 13 rose and fell with the day. A check in 32 other people, sampled at 12:45 and 4:15 in the afternoon, showed the same drift on 15 of the 17 clocks [3]. The tissue study compared 284 samples from 83 people aged 9 to 70: cheek swabs, saliva, dried blood spots and two kinds of blood cells. Most clocks were built on blood. Used on cheek cells or saliva, nearly all of them read older [4]. A 2026 analysis added meals. In 34 young adults sampled four times in one day, many clocks became only moderately or poorly repeatable after a single meal [5]. The last fix is to repeat the test. The largest shift in the two trials was a few months over three years, while a single reading can move by years between night and morning. So one reading can't yet show whether a new habit is working. Several readings, taken the same way, say more than one. The DO-HEALTH authors make the same point about their own trial: two time points carry more error than three or more [7]. So back to Maya's box. The science behind it is real. Her number will mean the most if she uses blood, tests at the same hour, and reads three results, not one. ## Our grade | Claim | Evidence | Grade | |---|---|---| | The newest clocks predict death and disease across large groups | 18,859 adults, 10 years, observational: 54% higher risk of death per standard step on GrimAge version 2, linked to 37 of 174 diseases [1] | Promising | | Changing your habits moves the score | Two randomized trials, small shifts: 2% to 3% on one clock [6], 3 to 4 months on two [7] | Promising | | One reading tells you your own pace of aging | The same sample read up to 9 years apart on older clocks [2]; about 3 years of swing by time of day in one man [3] | Weak | | A saliva or cheek-swab kit matches a blood test | 83 people: gaps of almost 30 years on some clocks built for blood [4] | Debunked | | A lower score means you will live longer | No trial has tested it | Unproven | | Overall: the newest clocks predict who gets sick and who dies sooner | One large 10-year study; a single home reading is still noisy | Promising | The case for the other side is that the proof is about groups, not about you. The big study watched people for 10 years. It did not test whether acting on a score helps anyone. The clock also added only a little to what ordinary facts already predicted, 0.851 to 0.865 [1]. A trial showing that people who lower their score go on to have fewer deaths or diseases would raise this grade. So would proof that one person's repeat home tests land close together. ## What this means for you The newest clocks are a real advance, and you can use one well. If you are curious and can spare the money, several hundred dollars a test [10], go in with the right picture. None of these home tests is approved by the FDA as a medical test [10]. Use a blood kit, not saliva. Test at the same time of day. Choose a lab that reports DunedinPACE or GrimAge. Wait for three or more results before you believe a trend. You also don't need the test to act. Your age, your weight and whether you smoke or drink already carried most of what the best clock could predict [1]. Don't change a habit because of one number. Stanford geneticist Anne Brunet said people "should not freak out if a clock says they're older than their calendar age," and that a younger reading "is to be taken with caution, too" [10]. Keep the numbers with the hardest proof in view: blood pressure, cholesterol, blood sugar and fitness. Attia's show notes for his episode carry this heading: "Why aging clocks are not yet reliable as consumer tools and why traditional health metrics still matter most" [8]. ## What's next The study that would settle this follows people whose clocks slowed in a trial and counts who gets sick. We'll update this page when one reports. For habits that changed health outcomes and not only a score, see [five health levers, checked one by one](/five-health-levers/) and [a few all-out bike sprints a week](/sprint-intervals/). For another case where a better blood number did not mean better health, read [the NAD booster trial](/nad-boosters/); for one where the test was the gold standard, see [NMN and aging muscle](/nmn/). And for another gadget that measures healthy people, see [continuous glucose monitors](/glucose-monitors/). Shopping for a test? Our [buyer's guide to biological age tests](/biological-age-test-guide/) grades 11 kinds, with prices by country. ## Sources 1. Mavrommatis C, Belsky DW, Ying K, Moqri M, Campbell A, Richmond A, Gladyshev VN, Chandra T, McCartney DL, Marioni RE. An unbiased comparison of 14 epigenetic clocks in relation to 174 incident disease outcomes. *Nature Communications*, 2025. 18,859 people, observational cohort (Generation Scotland), 10 years, 842 deaths. Funded by the Wellcome Trust and public bodies; one author invented DunedinPACE, licensed to TruDiagnostic. [doi:10.1038/s41467-025-66106-y](https://doi.org/10.1038/s41467-025-66106-y) 2. Higgins-Chen AT, Thrush KL, Wang Y, et al. A computational solution for bolstering reliability of epigenetic clocks: implications for clinical trials and longitudinal tracking. *Nature Aging*, 2022. 36 blood samples each run twice, laboratory reliability study. Two authors have clocks licensed to Elysium Health, which supplied some samples. [doi:10.1038/s43587-022-00248-2](https://doi.org/10.1038/s43587-022-00248-2) 3. Koncevičius K, Nair A, Šveikauskaitė A, Šeštokaitė A, Kazlauskaitė A, Dulskas A, Petronis A. Epigenetic age oscillates during the day. *Aging Cell*, 2024. One man sampled every 3 hours for 72 hours, plus 32 people sampled twice in one afternoon; observational. [doi:10.1111/acel.14170](https://doi.org/10.1111/acel.14170) 4. Apsley AT, Ye Q, Caspi A, Chiaro C, et al. Cross-tissue comparison of epigenetic aging clocks in humans. *Aging Cell*, 2025. 83 people, 284 samples from five tissues, observational. [doi:10.1111/acel.14451](https://doi.org/10.1111/acel.14451) 5. Sehgal R, et al. Biological Versus Technical Reliability of Epigenetic Clocks and Implications for Disease Prognosis and Intervention Response. *Aging Cell*, 2026. Reanalysis of eight repeat-sample datasets, 4 to 196 people each. Two authors have received consulting fees from TruDiagnostic. [doi:10.1111/acel.70635](https://doi.org/10.1111/acel.70635) 6. Waziry R, Ryan CP, Corcoran DL, et al. Effect of long-term caloric restriction on DNA methylation measures of biological aging in healthy adults from the CALERIE trial. *Nature Aging*, 2023. 197 people analyzed, randomized, 2 years. Two authors are inventors of DunedinPACE. [doi:10.1038/s43587-022-00357-y](https://doi.org/10.1038/s43587-022-00357-y) 7. Bischoff-Ferrari HA, Gängler S, Wieczorek M, Belsky DW, Ryan J, et al. 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. *Nature Aging*, 2025. 777 people, randomized, 3 years, analysis added after the trial. One author founded a foundation that licenses GrimAge. [doi:10.1038/s43587-024-00793-y](https://doi.org/10.1038/s43587-024-00793-y) 8. Attia P. #386: Aging clocks: what they measure, how they work, and their clinical and real-world relevance. *The Peter Attia Drive*, April 6, 2026. [peterattiamd.com/agingclocks](https://peterattiamd.com/agingclocks) 9. Johnson B. Post on X, February 19, 2025. [x.com/bryan_johnson/status/1892379570726457792](https://x.com/bryan_johnson/status/1892379570726457792) 10. Stanford Medicine. What biological age tests can and can't tell you. *Stanford Report*, September 23, 2026. [news.stanford.edu](https://news.stanford.edu/stories/2026/09/biological-age-tests-tips-advice) 11. Rejuvenation Olympics. DunedinPACE leaderboard, viewed October 12, 2026. [rejuvenationolympics.com/dunedin-pace](https://www.rejuvenationolympics.com/dunedin-pace)