Ageing Biomarkers & Epigenetic Clocks

Summary

Biological age estimation has moved from a research curiosity to a commercial industry. Epigenetic clocks are the best-known method, but functional measures like grip strength and VO2 max predict outcomes at least as well and cost far less. Individual-level measurement error is the field's central unresolved problem.

Last reviewed 2026-09-01

What it is

An ageing biomarker is a measurable parameter that predicts functional capacity or mortality risk better than chronological age alone, and ideally responds to interventions.

Horvath's multi-tissue methylation clock in 2013 was the breakthrough. Second-generation clocks — PhenoAge and GrimAge — were trained on mortality and clinical measures rather than chronological age and predict outcomes considerably better. DunedinPACE, released in 2022, estimates rate of ageing rather than accumulated age.

Statistical models fitted to large human cohort datasets, not direct measures of any single biological process.

A validated, responsive biomarker would allow longevity interventions to be tested in years rather than decades — the field's main practical bottleneck.

How it works

In plain terms

Chemical tags on DNA change in predictable patterns as we age. Statistical models read those patterns and estimate an age. Other approaches measure how well the body actually performs, which often predicts the future just as well.

Technical detail

First-generation clocks regress methylation at selected CpG sites against chronological age. Second-generation clocks such as GrimAge are trained on plasma protein surrogates and smoking pack-years against time-to-death, and outperform first-generation clocks for mortality prediction. DunedinPACE derives a pace-of-ageing estimate from longitudinal biomarker slopes in the Dunedin cohort. Proteomic and metabolomic organ-specific clocks are an active development area. Functional measures — gait speed, grip strength, VO2 max, one-leg stand — remain strongly predictive and highly reproducible.

Pathways involved

CpG methylation regression (Horvath, Hannum)
Mortality-trained composite clocks (PhenoAge, GrimAge)
Pace-of-ageing estimation (DunedinPACE)
Proteomic and organ-specific clocks
Functional capacity measures

Current research

Laboratory research

Array and sequencing platforms differ in the CpG sites they measure, producing systematic differences between clocks run on different platforms.

Animal research

Murine methylation clocks respond to caloric restriction, partial reprogramming and senolytics, providing a shorter path to intervention testing.

Human research

Epigenetic age acceleration predicts all-cause mortality across many cohorts, with GrimAge consistently strongest. The CALERIE caloric restriction trial reported a slowed DunedinPACE, one of the few randomised interventional signals. Test–retest reliability of first-generation clocks at individual level is poor — repeat measurements on the same sample can differ by several years — which is a serious limitation for consumer testing.

Ongoing research

Reliability improvement through principal-component-based clocks, and validation of organ-specific proteomic clocks.

What is being investigated

Epigenetic age acceleration as a mortality predictor
Pace-of-ageing measures as intervention endpoints
Organ-specific proteomic ageing clocks
Functional measures: grip strength, gait speed, VO2 max
Reliability and standardisation of consumer biological-age tests

These are research directions reported in the literature, not established effects or recommendations.

Risks, limitations and unknowns

Read this section before the rest

Individual-level test–retest error can span several years for first-generation clocks
Consumer biological-age tests often report precision the underlying method cannot support
Clocks are trained associations, not mechanistic measures — changing a clock reading is not proven to change outcomes
Platform and normalisation differences make results non-comparable between providers
No epigenetic clock is validated as a surrogate endpoint by any regulator

Evidence ratings

Population-level prediction

Strong

Robust mortality association across many cohorts.

Individual-level reliability

Limited

Test–retest error is substantial for several clocks.

Response to interventions

Preliminary

CALERIE signal is promising but isolated.

Functional measures

Strong

Grip strength and VO2 max are cheap and highly predictive.

References

[1]Horvath multi-tissue epigenetic clock — PubMed / Genome Biology, 2013
[2]GrimAge and mortality prediction — PubMed, 2019
[3]CALERIE and DunedinPACE — PubMed, 2023
[4]Reliability of epigenetic clocks — PubMed, 2022–present
Reference links open searches and records on PubMed and ClinicalTrials.gov so that every statement above can be traced to primary literature.

Frequently asked questions

What is biological age?+

An estimate of functional or physiological age derived from biomarkers, intended to predict outcomes better than years lived.

Are epigenetic clocks accurate?+

They predict mortality well at population level. At individual level, test–retest reliability of first-generation clocks is poor.

Which clock is most predictive?+

GrimAge and its successors, which were trained on mortality data rather than chronological age, consistently outperform first-generation clocks.

What is DunedinPACE?+

A measure of the rate of biological ageing derived from longitudinal data in the Dunedin birth cohort, rather than accumulated biological age.

Do consumer biological age tests work?+

They produce a number, but reported precision usually exceeds what the underlying method supports, and providers are not comparable to each other.

Can biological age be reduced?+

The CALERIE caloric restriction trial reported a slowed pace of ageing. Whether that translates into longer or healthier life has not been demonstrated.

What simple measures predict ageing well?+

Grip strength, gait speed, VO2 max and one-leg stand time are inexpensive, reproducible and strongly predictive of outcomes.

Are ageing clocks approved as clinical endpoints?+

No regulator has validated any epigenetic clock as a surrogate endpoint for trials.

Continue your research

Cellular health

Chronic inflammation

Epitalon research