Home / Reference / Longevity / Biological Age Tests

Test profile · DNA-methylation epigenetic clocks

Biological Age Tests

TruDiagnostic · Elysium Index · Horvath/GrimAge · DunedinPACE

Strong population-level signal DunedinPACE responsive in CALERIE Single-sample noise is real Cross-platform comparisons unreliable Not clinically actionable for one person

Epigenetic age clocks read patterns of DNA methylation — small chemical tags on CpG (cytosine-phosphate-guanine) sites across the genome — and compress those patterns into a single number meant to estimate biological versus chronological age. The science is real. The consumer use case is harder, because the noise on a single test can easily swamp the effect of a year of training or supplementation.

Substrate

DNA methylation

Sites measured

~850k CpGs

Major clocks

3 generations

Consumer price

$200–$500

01 / How they work

What an epigenetic clock is.

DNA does not change with age. Its methylation does. At specific CpG sites — places where a cytosine sits next to a guanine — a methyl group can be added or removed, switching the local epigenetic state. Steve Horvath's 2013 work showed that across 51 tissue types, methylation at 353 CpG sites could be combined into a linear model that predicted chronological age within a few years. That model is now called the Horvath clock [Horvath 2013].

Horvath was a first-generation clock — trained to estimate chronological age. Second-generation clocks (PhenoAge, GrimAge) were trained instead to predict morbidity and mortality, which gave them better association with health outcomes [Lu 2019]. DunedinPACE, published by Belsky and colleagues in 2022, is a different kind of measure entirely — instead of estimating an age, it estimates the rate at which someone is aging, derived from longitudinal data in the Dunedin birth cohort [Belsky 2022]. It is a speedometer, not an odometer.

02 / The clock generations

What each one actually measures.

Clock	Trained to predict	Generation	Note
Horvath 2013	Chronological age	First	353 CpGs, multi-tissue
Hannum 2013	Chronological age (blood)	First	71 CpGs, blood only
PhenoAge	Phenotypic age (mortality)	Second	Levine et al., 2018
GrimAge	Time to death	Second	Best mortality prediction
DunedinPACE	Pace of aging	Third (pace)	Speedometer, not age

Why DunedinPACE matters for interventions

First- and second-generation clocks are slow to move. They are integrated over many years of biology. A 6-month intervention will move them only if it produces a very large biological change. DunedinPACE was specifically designed to be sensitive to current rate, which is why it moved in the CALERIE caloric restriction trial when older clocks did not [Waziry 2023].

03 / What the trial data show

Things that actually move the clocks.

The CALERIE trial — the only randomized, controlled, multi-year trial of caloric restriction in healthy non-obese adults — reported that 25% caloric restriction for two years slowed DunedinPACE by 2–3% versus controls, while PhenoAge and GrimAge did not significantly differ [Waziry 2023]. A 2–3% slower pace of aging maps to roughly a 10–15% reduction in mortality hazard in observational data — small in any individual, meaningful at a population level.

Other interventions with positive signal in published or pre-print work include resistance and aerobic exercise (small effects), Mediterranean-style diet, multi-component lifestyle interventions, and — controversially — rapamycin in early observational work. See the rapamycin in humans piece for the longevity-drug context. The TAME (Targeting Aging with Metformin) trial, when it reads out, will include epigenetic age outcomes for the first large-scale geroprotector trial.

Things that do not reliably move the clocks (despite marketing claims): single-nutrient supplements without other lifestyle change, short-course IV therapies, and most "biohacker stacks" tested in n=1 designs.

04 / The reliability problem

Why single tests mislead.

Test-retest reliability for first-generation clocks (Horvath, Hannum) on the same blood sample, split between two aliquots, has been measured at roughly r = 0.85–0.95 — useful for population research, but corresponding to a noise band of ±2–4 years on any single biological age estimate [Higgins-Chen 2022]. That noise is wider than most interventions plausibly produce in six months.

DunedinPACE was deliberately engineered to be more reliable. The original paper reports intraclass correlation around 0.96 in repeated-measures datasets — the highest in the field [Belsky 2022]. Even so, single measurements come with a confidence interval that consumer reports rarely show.

Lab variance. Different sequencing runs, different lab batches, and even different hours of the day can shift the score.
Recent illness. Infection, vaccination, or surgery in the prior weeks can temporarily shift methylation patterns.
Cell-type composition. Blood is a mixture of cell types; shifts in lymphocyte vs neutrophil ratio can move scores independent of biology.
Cross-platform comparisons. A Horvath estimate from Lab A is not directly comparable to a Horvath estimate from Lab B. Sequential tracking should stay on the same platform.

05 / Consumer platforms

What you can actually buy.

Platform	Clocks reported	Approx price	Note
TruDiagnostic TruAge	Horvath, PhenoAge, DunedinPACE, OMICmAge	$200–$500	Most CpG sites covered consumer-side
Elysium Index	Proprietary clock	~$300	Distinct from public clocks
Glycanage	IgG glycan, not methylation	~$300	Different biology — inflammatory aging
InsideTracker	Blood-biomarker InnerAge	~$300	Blood chemistry, not epigenetics

06 / Bottom line

Where these tests are useful.

Epigenetic clocks are excellent population-level biomarkers and are appropriately used as endpoints in clinical trials. They are weaker as individual feedback tools: the noise floor is wide, the platforms are not interchangeable, and most published interventions move the score by less than the test-retest variance for first-generation clocks. DunedinPACE, with its higher reliability and better intervention responsiveness, is the most defensible single measure for someone who wants to track a longevity protocol over multiple years.

The honest framing: order one of these tests if you are curious, plan to repeat on the same platform every 12–24 months, and treat the result as one data point alongside cardiovascular biomarkers, body composition, and functional measures. Don't reorganize your life around a single year-on-year delta. See the longevity hub and the peptides 2026 overview for how aging biomarkers fit into broader longevity protocols.

07 / References

The evidence base.

Horvath S. DNA methylation age of human tissues and cell types. Genome Biology, 2013;14:R115. [Horvath 2013]
Hannum G, et al. Genome-wide methylation profiles reveal quantitative views of human aging rates. Molecular Cell, 2013;49(2):359–367. [Hannum 2013]
Levine ME, et al. An epigenetic biomarker of aging for lifespan and healthspan. Aging, 2018;10(4):573–591. [Levine 2018]
Lu AT, et al. DNA methylation GrimAge strongly predicts lifespan and healthspan. Aging, 2019;11(2):303–327. [Lu 2019]
Belsky DW, Caspi A, Corcoran DL, et al. DunedinPACE, a DNA methylation biomarker of the pace of aging. eLife, 2022;11:e73420. [Belsky 2022]
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;3:248–257. [Waziry 2023]
Higgins-Chen AT, et al. A computational solution for bolstering reliability of epigenetic clocks: implications for clinical trials and longitudinal tracking. Nature Aging, 2022;2:644–661. [Higgins-Chen 2022]
Fitzgerald KN, et al. Potential reversal of epigenetic age using a diet and lifestyle intervention. Aging, 2021;13(7):9419–9432. [Fitzgerald 2021]

About this profile

Last reviewed against evidence: 2026-05-12. This profile is editorial reference content, not sponsored. Wellness Radar does not carry affiliate links for biological age testing platforms and is not paid to recommend specific providers. Educational reference, not a diagnostic interpretation — clinical decisions belong with a clinician.