Epigenetic clocks track functional decline, at effect sizes nobody can use yet

Design

Longitudinal cohort analysis with linear mixed-effects models, testing twelve epigenetic clocks against functional trajectories

Population

4,018 Health and Retirement Study participants with DNA methylation data and up to 12 years of follow-up

Primary outcome

Trajectories of cognitive function, grip strength and walking speed by epigenetic clock, adjusted for chronological age

Effect

DNAmGrimAge: cognition beta -0.0141 (95% CI -0.0174 to -0.0107), grip -0.024 (-0.033 to -0.015), gait speed -0.0008 (-0.0013 to -0.0004), all p<0.001; Hannum cognition -0.0054 (-0.0095 to -0.0014)

Epigenetic clocks estimate biological age from DNA methylation, and the claim made for them is that they capture ageing that chronological age misses. This analysis tested that against the outcomes geriatricians actually measure: 4,018 Health and Retirement Study participants with methylation data and up to 12 years of follow-up, twelve different clocks, and longitudinal trajectories of cognition, grip strength and walking speed.

Adjusting for chronological age, older epigenetic age was associated with faster cognitive decline for the Hannum clock (beta -0.0054, 95% CI -0.0095 to -0.0014, p=0.009) and for DNAmGrimAge (-0.0141, -0.0174 to -0.0107, p<0.001). DNAmGrimAge was also associated with accelerated decline in grip strength (-0.024, -0.033 to -0.015) and walking speed (-0.0008, -0.0013 to -0.0004), both p<0.001. Of twelve clocks tested, one stood out consistently.

Two cautions. The analysis is described by its authors as retrospective, and the coefficients are per unit of epigenetic age per year - very small numbers whose clinical meaning is not established. An association that is statistically robust across 4,018 people and 12 years can still be too small to distinguish two patients in a clinic.

So this is a research finding, not a test. It matters because it narrows the field: if DNAmGrimAge outperforms eleven competitors on three separate functional domains, that is a signal about which biological processes matter, and it points future work at a specific clock rather than at the concept. Nobody should be ordering an epigenetic age, and commercial biological-age tests sold direct to patients remain unsupported by anything like this.

• Do not order or recommend commercial biological age testing - nothing here validates clinical use
• Grip strength and gait speed remain the measurable, modifiable markers to record in clinic
• Different epigenetic clocks are not interchangeable; most performed less well here
• Retrospective association with decline is not prediction of an individual's trajectory
• When a patient asks about a biological age result, discuss what actually changes function: activity, nutrition, medication review

The statistics, in plain English

These betas are interaction terms - the change in the rate of decline per unit of epigenetic age - which is why they look tiny: -0.024 for grip strength means a marginally steeper slope, not a measurable difference in kilograms at a clinic visit. Confidence intervals that exclude zero tell you the association is real in this dataset, not that it is large. A retrospective analysis of stored samples against subsequent decline is also vulnerable to whatever led to the sample being available, and the twelve clocks were tested together, so the standout result needs replication before it means much.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.