Epigenetic Clocks and Biological Age: How Accurate Are DNA Methylation Tests? (2026)

By Ryan Bethencourt, biotech investor and co-founder of IndieBio · Published September 23, 2025 · Last updated October 21, 2026 · Papers and test makers’ method pages checked October 21, 2026

In this article: What is an epigenetic clock? | How accurate are the tests? | The main clocks | What is DunedinPACE? | What each test measures | The TruMe test | TruDiagnostic vs Elysium Index | What clocks can and cannot tell you | Can epigenetic age be lowered? | What is not known | FAQ

What is an epigenetic clock?

An epigenetic clock is a formula that turns DNA methylation readings into an age estimate. Methylation is a small chemical tag on DNA, usually at sites where the base cytosine sits next to guanine (CpG sites). At some of these sites, the share of tagged DNA rises or falls with age in a predictable way (Bell et al., Genome Biology, 2019).

To build a clock, researchers measure methylation at hundreds of thousands of sites with an array, then let a regression model keep the subset of sites (71 in Hannum’s clock, 1,030 in GrimAge) whose combined readings best predict a target. The target is what separates the clocks:

• First-generation clocks, such as Horvath’s and Hannum’s, were trained to predict calendar age.
• Second-generation clocks, such as PhenoAge and GrimAge, were trained to predict a health measure: a mortality-risk score built from blood tests, or time to death.
• Pace-of-aging measures, such as DunedinPACE, were trained to predict how fast a person’s organ systems declined over two decades.

The output is an epigenetic age in years, which consumer tests usually report as biological age, or a rate for DunedinPACE. The gap between epigenetic age and calendar (chronological) age is called age acceleration. A positive gap means the clock reads older than expected for a person’s age; a negative gap means it reads younger.

How accurate are epigenetic age tests?

At estimating calendar age in research datasets, the best clocks are accurate: Horvath’s 2013 clock correlated with age at 0.96, with a median error of 3.6 years, in test data drawn from many tissues (Horvath, Genome Biology, 2013). A single commercial result is less certain, because lab noise, the tissue sampled and the measurement platform can each move it by years.

Lab noise

When the same blood sample is measured twice, a clock should return the same age. A 2022 study in Nature Aging found that technical noise produced differences of up to 9 years between such replicates for six prominent clocks. Rebuilding the clocks from principal components, a statistical summary of many sites at once, brought most replicates within 1.5 years of each other (Higgins-Chen et al., 2022).

DunedinPACE was built only from sites that measure reliably on repeat runs. On replicate samples its intraclass correlation was 0.96, where 1.0 means perfect agreement; its authors found it as reliable as GrimAge and more reliable than the other clocks they compared (Belsky et al., eLife, 2022).

Saliva, cheek cells and blood

Most clocks were trained on blood. A 2025 study in Aging Cell compared 284 samples of five tissue types from 83 people aged 9 to 70. It found large differences within the same person between oral samples (cheek cells and saliva) and blood samples, with average differences of almost 30 years for some clocks. The Skin and Blood clock agreed best across tissues (Apsley et al., 2025).

A second 2025 analysis pooled 107 people with both saliva and blood data. After correcting for cell types, agreement between the two tissues was moderate for newer measures (intraclass correlations of 0.76 for the principal-component version of GrimAge, 0.72 for that of PhenoAge, and 0.68 for DunedinPACE) and poor, 0.19 to 0.25, for first-generation age acceleration. The authors concluded that this level of agreement may not suit clinical or commercial use (Zarandooz and Raffington, Clinical Epigenetics, 2025).

The lab platform

Clocks are tied to the array they were built on. Illumina’s newer EPICv2 array dropped some of the sites older clocks rely on, and a 2024 study found that four existing clocks run on the EPICv2 probe set were distorted, with average differences of up to 25 years; the authors then built a new model that works on the older and newer arrays (Garma and Quintela-Fandino, Genome Medicine, 2024).

What this means for a result. A gap of a year or two between epigenetic age and calendar age is within the error of most clocks. Two companies, or two clocks in one report, can give the same person different numbers, and those numbers are not interchangeable.

What are the main epigenetic clocks, and how do they differ?

The table compares six clocks: Horvath and Hannum (built to match calendar age), PhenoAge and GrimAge (built to predict health and death), DunedinPACE (a speed of aging) and, for cheek swabs, CheekAge. They differ in the target they were trained on, the number of CpG sites and the tissue they were built in.

Clock (year)	Trained to predict	CpG sites	Built from	Published performance
Horvath pan-tissue (2013)	Calendar age	353	About 8,000 samples from 51 healthy tissues and cell types	Correlation with age 0.96, median error 3.6 years in test data
Hannum (2013)	Calendar age	71	Whole blood from 656 people aged 19 to 101, measured at more than 450,000 sites	Correlation 0.96, error 3.9 years; 0.91 and 4.9 years in a validation group
PhenoAge (2018)	Phenotypic age: a score from nine clinical blood markers plus calendar age	513, selected from 20,169 candidates	Whole blood from 456 people in the InCHIANTI study, at two time points	Correlation with age 0.62 to 0.89 across five validation groups; each extra year linked to a 4.5 percent higher risk of death
GrimAge (2019)	Time to death, through methylation estimates of seven blood proteins and smoking pack-years	1,030	Blood from 2,356 people in the Framingham Heart Study Offspring Cohort (1,731 for training)	Best time-to-death prediction of the clocks compared, in 6,935 people; each extra year of acceleration linked to about 10 percent higher risk of death
DunedinPACE (2022)	Pace of aging: yearly decline across 19 markers of organ function, measured at ages 26, 32, 38 and 45	173	Blood taken at age 45 from 817 members of the Dunedin Study, a cohort of 1,037 people born in 1972 and 1973	Repeat-measurement reliability 0.96; evaluated in five more datasets, where it was linked to illness, disability and death
CheekAge (2024)	Calendar age, with lifestyle and health factors built into training	Built from about 200,000 filtered sites, grouped into clusters	Cheek swabs from 8,045 adults	In blood from two Scottish birth cohorts, each standard deviation linked to 21 percent higher risk of death

GrimAge has a second version. GrimAge2, published in 2022, adds methylation estimates of C-reactive protein and hemoglobin A1C, and in 13,399 blood samples from nine cohorts it predicted death better than the original (Lu et al., Aging, 2022).

The risk figures in the table are averages across thousands of people. PhenoAge’s 4.5 percent and GrimAge’s hazard ratio of 1.10 per extra year describe groups, not the future of any one person.

What is DunedinPACE?

DunedinPACE is a blood test that estimates how fast a person is aging now, in years of biological decline per calendar year, rather than how old the person is. A score of 1.0 is the average pace, and a score above 1.0 means faster than average.

It comes from the Dunedin Study, which has followed a birth cohort of 1,037 people born in 1972 and 1973 in Dunedin, New Zealand. Researchers measured 19 markers of cardiovascular, metabolic, kidney, liver, immune, dental and lung function at ages 26, 32, 38 and 45, calculated each person’s rate of decline, and, using blood taken at age 45 from 817 study members, trained a 173-site methylation model to predict that rate from a single blood sample (Belsky et al., eLife, 2022).

In the paper’s validation datasets, a faster DunedinPACE was associated with illness, disability and death, with effect sizes similar to GrimAge, and it added prediction beyond GrimAge. It is also the measure that moved in the CALERIE calorie-restriction trial, described below.

What does each biological age test measure?

Consumer tests differ in the sample, the number of sites the lab reads and the algorithm that turns the readings into an age, so their results cannot be compared with each other. The table summarizes what each maker’s method page and the published papers say, as of October 21, 2026. Prices are left out because they change often.

Test (maker)	Sample	What the lab reads	Clock or algorithm	What the report gives	Maker’s method page
TruAge (TruMe Labs)	Saliva	Published papers: methylation at 9 CpG sites, by Sanger sequencing or pyrosequencing. TruMe’s website: “over 850,000 specific sites”	Proprietary 9-site regression clock in the papers; the website names no published clock	Biological age score, an aging-rate analysis and lifestyle recommendations	TruMe science page
TruAge (TruDiagnostic)	Finger-stick blood	Genome-wide methylation array; TruDiagnostic says it reads more than 1 million sites	OMICmAge (biological age), DunedinPACE (pace of aging), SYMPHONYAge (11 organ systems)	Biological age, pace of aging and organ-system ages, all estimated from methylation	About TruAge
Index (Elysium Health)	Saliva	A custom “Elysium chip” made with Illumina; the number of sites is not stated	APEX, Elysium’s platform; Elysium says it partnered with Morgan Levine of Yale and cites her lab’s principal-component method (Higgins-Chen et al., 2022); that paper discloses that aging metrics built with the method are licensed by Elysium through Yale University	Biological age, cumulative rate of aging and ages for nine body systems	Science behind Index
TallyAge (Tally Health)	Cheek swab	“Over 200,000 sites across the entire genome” (Tally Health)	TallyAge, which Tally Health says was inspired by its CheekAge clock, trained on cheek swabs from 8,045 adults	TallyAge result plus a lifestyle plan	Tally Health science page
GlycanAge	Finger-prick blood	Sugar chains (glycans) on the antibody immunoglobulin G; not DNA methylation	Glycan-based age index; the age-related glycan changes are described in Krištić et al. (2014)	A glycan-based biological age	GlycanAge science page

Several makers report extra scores, such as organ-system ages. These are methylation-based estimates trained against other measurements, not direct tests of the organs.

Two makers state plainly that their reports are not diagnostic. Elysium says Index has not been reviewed or approved by FDA because it is not a medical device. TruDiagnostic says its results should not be the sole basis for diagnosis or treatment decisions.

What does the TruMe DNA methylation test measure?

TruMe’s test, sold by TruMe Labs as TruAge, estimates biological age from a saliva sample. In the peer-reviewed papers that used it, the clock reads methylation at 9 CpG sites by Sanger sequencing or pyrosequencing and combines the readings in a proprietary regression formula (Pabis et al., Aging Cell, 2026).

According to that 2026 paper, the clock’s developers screened more than 900 age-related CpG sites from public datasets and kept nine in regulatory regions of aging-associated genes. Among 4,260 people who took the test between 2020 and 2025, the clock’s mean absolute error against calendar age was 5.4 years. In an earlier group of 42 people, the median absolute error was 4.23 years (Demidenko et al., Aging, 2021). A 2025 paper in npj Aging describes the same nine-site clock, read by Sanger sequencing, and a broader TruAge Index built from about 33 sites with the nine at its core (Dakhovnik et al., 2025). All three papers list TruMe staff among the authors.

TruMe’s website, read on October 21, 2026, describes something larger: analysis of more than 850,000 methylation sites with a proprietary algorithm. The site does not name a published clock or say whether this method replaced the 9-site clock in the papers, and a PubMed search on October 21, 2026 found no paper describing an 850,000-site TruMe clock.

The name causes confusion. TruDiagnostic, a separate company in Lexington, Kentucky, also sells a test called TruAge, based on a finger-stick blood sample and a genome-wide array, and it reports DunedinPACE among other measures. The two products share a name but not a method.

TruDiagnostic vs Elysium Index: what is the difference?

TruDiagnostic’s TruAge reads a finger-stick blood sample on a genome-wide array and reports named clocks, including the published DunedinPACE. Elysium’s Index reads saliva on a custom chip and reports its own biological age, rate of aging and nine system ages, and cites the principal-component clock method of Higgins-Chen et al. (2022).

As of October 21, 2026	TruDiagnostic TruAge	Elysium Index
Sample	Finger-stick blood, collected at home	Saliva, collected at home
Lab readout	Genome-wide methylation array (more than 1 million sites, per TruDiagnostic)	Custom Elysium chip made with Illumina (site count not stated)
Algorithms named	OMICmAge, DunedinPACE, SYMPHONYAge	APEX, Elysium’s platform; Elysium cites the principal-component method of Higgins-Chen et al. (2022)
Main outputs	Biological age, pace of aging, ages for 11 organ systems	Biological age, cumulative rate of aging, ages for nine body systems
Reliability claim	TruDiagnostic’s page states an average intraclass correlation above 0.98 without naming the study; the DunedinPACE paper reported 0.96 on replicates (Belsky et al., 2022)	Elysium says Index agrees across replicates within zero to 1.5 years and cites Higgins-Chen et al. (2022); that paper discloses that aging metrics built with its method are licensed by Elysium through Yale University and that Elysium supplied replicate data for it
Maker’s stated use	“Not a diagnostic test”; an adjunct to clinical evaluation	Not reviewed or approved by FDA; “not a medical device”

The tissue matters as much as the brand. Blood-trained clocks fit blood. Saliva results depend on how well the algorithm was adapted to saliva’s different mix of cells. Elysium says it tested its method under saliva conditions, but the cross-tissue studies above show why a saliva number and a blood number for the same person should not be compared directly.

What can an epigenetic clock tell you, and what can’t it?

Across large groups, clocks that read older than expected predict more disease and earlier death. For one person, a clock gives a rough, noisy estimate that cannot diagnose a disease, and it cannot show on its own that a product or habit works.

What clocks can tell you

• Risk across a population. In five validation cohorts, each extra year of PhenoAge beyond calendar age was linked to a 4.5 percent higher risk of death (Levine et al., 2018). In 6,935 people, each extra year of GrimAge acceleration carried a hazard ratio for death of 1.10, the strongest of the clocks compared (Lu et al., 2019).
• Change in a trial group. Averaged over hundreds of people, a reliable measure can show whether an intervention shifted biology, as DunedinPACE did in the CALERIE trial.
• A research signal in cells. Horvath’s clock reads close to zero in embryonic and induced pluripotent stem cells (Horvath, 2013), which makes clocks a natural readout for partial reprogramming research.

What clocks cannot tell you

• A diagnosis. No epigenetic clock is a diagnostic test, and Elysium and TruDiagnostic say so about their own tests.
• An exact personal forecast. Hazard ratios describe averages across thousands of people.
• Proof that something works. A 2024 review in Nature Medicine found no consensus on how biomarkers of aging should be validated and said their clinical usefulness still has to be shown in prospective trials (Moqri et al., 2024).
• A fair comparison across tests. Different clocks, tissues and lab platforms give the same person different numbers.
• Meaning in a small change. A shift of a year or two between tests can come from lab noise alone.

Can you lower your epigenetic age?

Randomized trials such as CALERIE and DO-HEALTH have shifted some clocks by small amounts, and no trial has yet shown that lowering a clock score leads to less disease or a longer life. Larger reported reversals come from small studies: an 8-week pilot randomized trial in 43 men (Fitzgerald et al., 2021) and uncontrolled studies such as TRIIM and the Rejuvant review.

Study	Design and participants	What changed	Limits
CALERIE (Waziry et al., Nature Aging, 2023)	Randomized trial, post hoc analysis: 220 adults without obesity assigned to a 25 percent calorie-restriction target or their usual diet for 2 years; 197 had methylation data for the analysis	DunedinPACE slowed by 2 to 3 percent	PhenoAge and GrimAge did not change significantly; effects were small
DO-HEALTH (Bischoff-Ferrari et al., Nature Aging, 2025)	Randomized factorial trial, post hoc analysis: 777 of 2,157 generally healthy adults aged 70 and older, all from the Swiss sites; the trial tested vitamin D (2,000 IU a day), omega-3 (1 g a day) and a home exercise program, alone and in combination, for 3 years	Omega-3 slowed PhenoAge, GrimAge2 and DunedinPACE; the PhenoAge and GrimAge2 reductions were 2.9 to 3.8 months over 3 years and the DunedinPACE effect about 1 percent; the three treatments added up on PhenoAge	The authors called the effects small; the Swiss participants were healthier and more active than the trial as a whole
Diet and lifestyle pilot (Fitzgerald et al., Aging, 2021)	Pilot randomized trial: 43 healthy men aged 50 to 72; an 8-week program of diet, sleep, exercise and relaxation guidance with probiotic and phytonutrient supplements, compared with no intervention	Horvath (2013) epigenetic age, measured in saliva, fell by 3.23 years relative to the control group; the 18 treated men analyzed averaged 2.04 years below their own baseline (figures as corrected in 2024)	Small and short, with one first-generation clock read in saliva; the authors called for larger and longer trials. Funded by an unrestricted grant from Metagenics, Inc.; two authors disclosed that they use the program in clinical practice, are named on a related patent application and earn from related educational products
TRIIM (Fahy et al., Aging Cell, 2019)	One-year pilot: 9 men aged 51 to 65 received growth hormone with DHEA and metformin	Mean epigenetic age about 1.5 years below baseline after a year; GrimAge about 2 years lower relative to calendar age	Nine volunteers and no control group; the authors called for an appropriately powered follow-up
Rejuvant review (Demidenko et al., Aging, 2021)	Retrospective analysis: 42 people who chose to take an alpha-ketoglutarate formulation for 4 to 10 months, measured with the TruMe saliva test	Average drop of about 8 years in TruMe biological age	Not placebo-controlled, the authors’ stated main limitation; the data came from users of Ponce de Leon Health’s Rejuvant product, TruMe had a testing agreement for those customers, and authors included staff of TruMe and Ponce de Leon Health

The DO-HEALTH authors noted that there is no gold standard measure of biological aging. The CALERIE authors wrote that a conclusive test of whether slowing aging biology helps people will need trials with long-term follow-up for outcomes such as chronic disease and death (Waziry et al., 2023). Clocks do not always fall after a treatment aimed at aging, either: in a 2026 study, three new clocks built from senescence-related methylation sites stayed the same or even accelerated after senolytic treatment, in cultured cells and in blood from human senolytic trials, and the authors noted that two earlier human trials of dasatinib and quercetin, with or without fisetin, found no change in many existing clocks (Kasamoto et al., Aging Cell, 2026). Our guide to senolytics such as dasatinib, quercetin and fisetin covers those drugs. For more on the TRIIM pilot, see Resetting Biological Age: Insights from the TRIIM Trial.

What is still not known about epigenetic clocks?

• Cause or marker. No trial has yet shown that slowing a clock leads to fewer diseases or deaths, so it is not known whether the clocks measure something that drives aging or only track it.
• What the changes do. The biology behind age-related methylation changes is still being worked out (Bell et al., 2019).
• Saliva versus blood. Researchers have called for paired blood and saliva samples from the same people to validate saliva versions of blood clocks (Zarandooz and Raffington, 2025).
• Independent checks of consumer algorithms. Several consumer clocks have been tested mainly in papers co-written by the companies’ own scientists, including CheekAge (Tally Health) and TruAge (TruMe). Published methods behind other tests carry commercial ties too: the principal-component clock paper discloses that aging metrics built with its method are licensed by Elysium through Yale University (Higgins-Chen et al., 2022), and six DunedinPACE authors disclosed that they are inventors on a Duke University and University of Otago invention licensed to a commercial entity (Belsky et al., 2022). TruDiagnostic’s page, read October 21, 2026, calls DunedinPACE “our aging speed clock developed with Duke”.

Frequently asked questions

How accurate are epigenetic age tests?

Against calendar age, the best research clocks correlate at about 0.96, with a median error of about 3.6 years (Horvath, 2013). Single consumer results are less precise: repeat runs of the same sample have differed by up to 9 years on the original versions of six well-known clocks, and saliva or cheek results can differ widely from blood. A difference of a year or two between epigenetic age and calendar age is within the error of most clocks.

What is the best biological age test?

It depends on the question, because the tests measure different things. For predicting death and disease in research, GrimAge and its 2022 update were validated on thousands of blood samples and predicted death better than earlier clocks, and DunedinPACE is as reliable on repeat measurement as GrimAge and more reliable than the other clocks its authors compared. For saliva or cheek swabs, a clock built for that tissue is a better match than a blood clock. A test whose clock and validation are published can be checked; a proprietary one cannot.

How many CpG sites do PhenoAge and GrimAge use?

PhenoAge uses 513 CpG sites, selected from 20,169 candidate sites (Levine et al., 2018). GrimAge uses 1,030 unique CpG sites, which feed methylation estimates of seven blood proteins and of smoking pack-years (Lu et al., 2019). For comparison, Horvath’s 2013 clock uses 353 sites, Hannum’s 71 and DunedinPACE 173.

What does the TruMe DNA methylation test measure?

It estimates biological age from saliva. Peer-reviewed papers that used the test describe a proprietary clock that reads 9 CpG sites by Sanger sequencing or pyrosequencing, with a mean absolute error of 5.4 years against calendar age among 4,260 people who took the test (Pabis et al., Aging Cell, 2026). TruMe’s website, read on October 21, 2026, describes analysis of more than 850,000 methylation sites but does not name a published clock.

Is TruMe’s TruAge the same as TruDiagnostic’s TruAge?

No. They are tests of the same name from different companies. In the published papers, TruMe’s TruAge uses saliva and a 9-site clock. As of October 21, 2026, TruDiagnostic’s TruAge uses a finger-stick blood sample read on a genome-wide array and reports OMICmAge, DunedinPACE and SYMPHONYAge.

Is a saliva test as good as a blood test for epigenetic age?

The two are not interchangeable. Most clocks were built in blood, and a 2025 study of 83 people found average within-person differences of almost 30 years between oral and blood samples for some clocks. Clocks built for oral samples, such as CheekAge for cheek swabs, exist for this reason, but saliva and blood results for the same person should not be compared directly.

Can an epigenetic age test diagnose disease?

No. Clocks predict risk across large groups, not disease in one person. As of October 21, 2026, Elysium says its Index test is not a medical device and has not been reviewed by FDA, and TruDiagnostic says its results should not be the sole basis for diagnosis or treatment decisions. Questions about a result are best taken to a clinician who knows the person’s medical history.

Can you lower your epigenetic age?

Post hoc analyses of two larger randomized trials found small changes. Calorie restriction slowed DunedinPACE by 2 to 3 percent over two years in the CALERIE trial of adults without obesity, and omega-3 slowed three clocks over three years among 777 DO-HEALTH participants aged 70 and older, by 2.9 to 3.8 months on PhenoAge and GrimAge2. An 8-week pilot trial in 43 men reported a 3.23-year drop relative to controls on Horvath’s clock, measured in saliva (Fitzgerald et al., 2021). No trial has shown that lowering a clock score reduces disease or extends life.

Sources

1. Horvath S. “DNA methylation age of human tissues and cell types.” Genome Biology, 2013;14(10):R115. PMC4015143
2. Hannum G, et al. “Genome-wide methylation profiles reveal quantitative views of human aging rates.” Molecular Cell, 2013;49(2):359-367. PMC3780611
3. Levine ME, et al. “An epigenetic biomarker of aging for lifespan and healthspan.” Aging, 2018;10(4):573-591. PMC5940111
4. Lu AT, et al. “DNA methylation GrimAge strongly predicts lifespan and healthspan.” Aging, 2019;11(2):303-327. PMC6366976
5. Lu AT, et al. “DNA methylation GrimAge version 2.” Aging, 2022;14:9484-9549. PMC9792204
6. Belsky DW, et al. “DunedinPACE, a DNA methylation biomarker of the pace of aging.” eLife, 2022;11:e73420. PMC8853656
7. Higgins-Chen AT, et al. “A computational solution for bolstering reliability of epigenetic clocks: implications for clinical trials and longitudinal tracking.” Nature Aging, 2022;2(7):644-661. PMC9586209
8. Garma LD, Quintela-Fandino M. “Applicability of epigenetic age models to next-generation methylation arrays.” Genome Medicine, 2024;16:116. PMC11460231
9. Apsley AT, et al. “Cross-tissue comparison of epigenetic aging clocks in humans.” Aging Cell, 2025;24(4):e14451. PMC11984668
10. Zarandooz S, Raffington L. “Applying blood-derived epigenetic algorithms to saliva: cross-tissue similarity of DNA-methylation indices of aging, physiology, and cognition.” Clinical Epigenetics, 2025;17:61. PMC12016411
11. Shokhirev MN, et al. “CheekAge: a next-generation buccal epigenetic aging clock associated with lifestyle and health.” GeroScience, 2024;46(3):3429-3443. PMC11009193
12. Shokhirev MN, et al. “CheekAge, a next-generation epigenetic buccal clock, is predictive of mortality in human blood.” Frontiers in Aging, 2024;5:1460360. PMC11473594
13. Pabis K, et al. “Supplements and Drugs Are Associated With Biological Age in a Cohort of Exceptionally Healthy Individuals.” Aging Cell, 2026;25(6):e70517. PMC13239866
14. Demidenko O, et al. “Rejuvant, a potential life-extending compound formulation with alpha-ketoglutarate and vitamins, conferred an average 8 year reduction in biological aging, after an average of 7 months of use, in the TruAge DNA methylation test.” Aging, 2021;13(22):24485-24499. PMC8660611
15. Dakhovnik A, et al. “A collagen amino acid composition supplementation reduces biological age in humans and increases health and lifespan in vivo.” npj Aging, 2025;11:91. PMC12635253
16. Krištić J, et al. “Glycans are a novel biomarker of chronological and biological ages.” Journals of Gerontology Series A, 2014;69(7):779. PMC4049143
17. Waziry R, 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;3:248-257. PMC10148951
18. Bischoff-Ferrari HA, 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;5(3):376-385. PMC11922767
19. Fahy GM, et al. “Reversal of epigenetic aging and immunosenescent trends in humans.” Aging Cell, 2019;18(6):e13028. PMC6826138
20. Fitzgerald KN, et al. “Potential reversal of epigenetic age using a diet and lifestyle intervention: a pilot randomized clinical trial.” Aging, 2021;13(7):9419-9432; correction in Aging, 2024;16(5):4943-4945. PMC8064200
21. Kasamoto J, et al. “DNA methylation signatures of cellular senescence are not reversed by senolytic treatment.” Aging Cell, 2026;25(3):e70430. PMC12938503
22. Moqri M, et al. “Validation of biomarkers of aging.” Nature Medicine, 2024;30(2):360-372. PMC11090477
23. Bell CG, et al. “DNA methylation aging clocks: challenges and recommendations.” Genome Biology, 2019;20:249. PMC6876109
24. Test makers’ method pages, read October 21, 2026: TruMe Labs; TruDiagnostic; Elysium Health; Tally Health; GlycanAge.

More on Grey Area Labs

• Longevity science: the Grey Area Labs hub
• Resetting biological age: the TRIIM trial
• Senolytics: dasatinib, quercetin and fisetin
• Longevity clinical trials
• Grey hair as a clue to biological age

Updated October 21, 2026: rewritten with each clock’s site count checked against its paper (PhenoAge 513, GrimAge 1,030), correlations in place of R squared values, new evidence on lab noise, saliva and trials, and a table of what the TruMe, TruDiagnostic, Elysium, Tally Health and GlycanAge tests measure, absorbing the February 2026 test comparison.

This article explains published research. It is not medical advice. How we write about health.