Epigenetic age test (biological age)

What the test is and what it promises

Small chemical tags, methyl groups, sit at certain sites on the DNA. Their pattern shifts with age in a characteristic way. Epigenetic clocks are machine-learning models that estimate a number from this pattern. Which number depends on what the clock was trained on: the first generation, such as the Horvath and Hannum clocks, on chronological age, the next, such as PhenoAge and GrimAge, on health markers and time to death, DunedinPACE on the pace at which organ functions decline over the years (Apsley 2025). How epigenetics works in principle and how prevention services use it is covered under Prevention, Epigenetics & Genetics; genetic tests, by contrast, examine the unchangeable gene sequence.

The tests are offered online, with a saliva sample sent in by post, or through practices with a blood draw. Providers use different algorithms and measure different methylation sites (Apotheken Umschau 2023). The promise is understandable: a tangible number for where you stand, and a second measurement that shows whether your changes in diet, exercise or sleep have had an effect. Programs that sell exactly this before-and-after comparison are described on the page Epigenetic coaching.

What the studies show

Faul 2023 – prediction in a representative sample

Analysis of the Health and Retirement Study with 3,581 older Americans. Age acceleration by PhenoAge, GrimAge and DunedinPACE was consistently a significant predictor of cognitive impairment, functional limitations and chronic disease two years later and of mortality over four years. Lower-noise versions based on principal components did not substantially change these associations. The authors consider the value as a predictor clear, but emphasize: demographics, socioeconomic status, mental health and health behaviors predicted later outcomes at least as well.

Higgins-Chen 2022 – technical measurement noise

When the same sample was measured several times, six widely used clocks deviated from one another by up to 9 years from technical noise alone. The research group retrained the clocks on principal components of the methylation data. These versions agreed within 1.5 years for most replicates, detected intervention effects better and produced stable trajectories.

Sehgal 2026 – technical versus biological reliability

The study examined 18 methylation clocks. Technically, most were highly reproducible, but some reacted to laboratory factors such as the position on the measurement chip or the DNA extraction method; PCGrimAge and SystemsAge remained robust. Biological reliability was markedly weaker: on repeat measurements at short intervals, under influences such as meals, stress and environmental exposures, most clocks were only poorly to moderately stable. Less reliable clocks produced variable or misleading results in interventions.

Zarandooz and Raffington 2025 – saliva versus blood

Three datasets with saliva and blood samples from the same 107 people aged 5 to 74. After correcting for cell composition, newer measures in saliva agreed moderately with the blood value: PCGrimAge with an intraclass correlation coefficient of 0.76, PCPhenoAge 0.72, DunedinPACE 0.68. For age acceleration by first-generation clocks, agreement was only 0.19 to 0.25. This may be enough for some research questions, the authors write, but possibly not for clinical or commercial applications.

Bruellman 2026 – saliva and blood in young adults

91 adults, 31 years old on average, gave saliva and two types of blood samples. Across 15 clocks, methylation age in saliva was 3.77 to 19.72 years higher than in blood, while the two blood fractions were comparable. One exception was DunedinPACE, which came out similar in all sample types. According to these data, saliva is not a direct substitute for blood, and comparisons only make sense within the same sample type.

CALERIE 2023 – caloric restriction as an endpoint study

220 adults without obesity were randomized for two years to 25 percent caloric restriction or a normal diet; the methylation clocks were analyzed retrospectively. DunedinPACE showed a 2 to 3 percent slower pace of aging, while PhenoAge and GrimAge did not change significantly. The authors see this as support for the idea that aging can be influenced biologically, and call for trials with disease and mortality as endpoints for definitive proof (Waziry 2023). Background on fasting and calorie reduction is available under Fasting & autophagy.

TRIIM 2019 – small study without a control group

Nine healthy men aged 51 to 65 received a combination of growth hormone, DHEA and metformin for one year, originally with the aim of regenerating the thymus. According to four clocks, epigenetic age after one year was on average about 1.5 years below baseline, or by calculation 2.5 years below the value expected without treatment; by GrimAge, a gap of 2 years persisted six months after the end of treatment (Fahy 2019). The study is often cited as evidence of “rejuvenation”. With nine participants and no control group, it is a hint, not proof, and the active substances are prescription drugs with their own risks, see Growth hormone, DHEA and Metformin. Further intervention studies, for example with omega-3, are presented on the Epigenetic coaching page.

What users report

In the European iHelp study, 178 adults from a charitable community testing program received a personal report with epigenetic age, genetic cancer risk scores and lifestyle data. Of the 91 who answered a survey after one year, 72.5 percent said they felt healthier, 60.4 percent were more physically active, 47.3 percent had improved their diet and 63.7 percent had learned more about health (Yang 2026). These are self-reports without a control group, answered by just over half of the participants, not proof of efficacy. A side finding of the same study is notable: 14 clocks compared gave clearly different results for the same samples.

Where the data stop

The strength of the clocks lies in the average across many people. For the individual, the uncertainties add up: technical noise, day-to-day fluctuations, differences between saliva and blood, between clocks and between laboratories. A change of one or two years after a program can hardly be separated from this. A 2025 review concludes that epigenetic clocks do not meet common standards for clinical utility compared with established biomarkers, and that individual-level decisions can be uninformative and potentially harmful (Apsley 2025).

Above all, the decisive proof is missing: no study has tested whether people who take such a test and act on it stay healthier or live longer than people without a test. Nor has it been shown that a lowered clock value prevents disease; the CALERIE authors themselves call for long trials with hard endpoints. And because income, mental health and behavior predict the course at least as well (Faul 2023), the number often provides less new information than its price suggests.

Status, approval and legal

Epigenetic age tests are offered in Germany as a self-pay service, online or through practices; according to a report in Apotheken Umschau, prices start at around 200 euros, with no fixed upper limit. There is no guideline that recommends the test for healthy people, and experts in aging research view the commercial offerings critically. The statutory health check-up, which is covered by German statutory health insurance, is described under Prevention, Epigenetics & Genetics.

The sample contains your DNA. Before sending it in, ask which laboratory analyzes it, how long the sample and raw data are stored and whether they are passed on to third parties.

Safety

Physically, the test is low-risk; it only needs a saliva or blood sample. The risks lie in the interpretation: an unfavorable value can cause unnecessary worry, a favorable one can falsely reassure and tempt people to neglect proven prevention such as blood pressure and cholesterol checks or cancer screening. Caution is also warranted when a single result leads directly to a supplement package.

Anyone who wants to test should ask beforehand which clock is used, whether saliva or blood is analyzed and how large the measurement uncertainty is. Before-and-after values are only comparable in the same laboratory, with the same clock and the same sample material.

Sources

• Faul JD et al., Proc Natl Acad Sci U S A 2023 – Health and Retirement Study, 3,581 older Americans, prediction of disease and mortality (PMID 36802439)
• Higgins-Chen AT et al., Nat Aging 2022 – technical reliability of epigenetic clocks, principal-component versions (PMID 36277076)
• Sehgal R et al., Aging Cell 2026 – technical and biological reliability of 18 methylation clocks (PMID 42525215)
• Zarandooz S, Raffington L, Clin Epigenetics 2025 – blood algorithms applied to saliva, 107 people with both samples (PMID 40270051)
• Bruellman R et al., Epigenetics 2026 – methylation clocks in saliva and blood, 91 young adults (PMID 41985071)
• Waziry R et al., Nat Aging 2023 – CALERIE, caloric restriction and methylation clocks, randomized (PMID 37118425)
• Fahy GM et al., Aging Cell 2019 – TRIIM, 9 men, no control group (PMID 31496122)
• Apsley AT et al., Epigenomics 2025 – limits of epigenetic clocks as personal biomarkers (PMID 41403206)
• Yang NY et al., Epigenomes 2026 – iHelp, feedback of epigenetic age, survey after one year (PMID 42029636)
• Apotheken Umschau, September 14, 2023 – Jung bleiben: Wie alt bin ich wirklich? (Staying young: how old am I really?) Report on epigenetic age tests with price information