Biological age: what epigenetic clocks really measure

Your birth date says little about how well your heart, vessels, muscles and metabolism are actually doing — two 50-year-olds can be worlds apart biologically. "Biological age" is the attempt to put that state into one number. The problem: there is no unified definition. Depending on whether you use blood values, fitness, grip strength or molecular markers, a different number comes out.

The most prominent variant are epigenetic clocks: with age, chemical marks on the DNA (methyl groups) that switch genes on and off change. These patterns are so regular that algorithms can estimate an age from them — Steve Horvath's 2013 clock did so across the most diverse tissues with an error of only a few years. That was a genuine scientific breakthrough and remains the foundation of the field.

The first clock generation (Horvath, Hannum) was trained to hit calendar age — fascinating, but of limited use for health questions: if you know your birth date, you don't need a test for it. Things got interesting with the second generation: GrimAge was trained on mortality and smoking data and predicts lifespan and disease in cohorts clearly better than a birth date. DunedinPACE takes a different route and estimates the current pace of aging from a birth cohort followed for decades — how many "biological years" pass per calendar year.

Important for perspective: these achievements are group statistics. That a clock separates risks in a cohort of thousands does not mean a single person's measurement makes a reliable statement about their future — just as a population-level risk factor can be real and still not be destiny for an individual.

This is exactly where the booming test market ("mail in a saliva sample, learn your biological age") gets shaky. First, the technical noise: analyses show that the same sample can deviate by several years (up to nine in the extreme) on repeat measurement depending on the clock. Researchers developed computational corrections (PC clocks) specifically for this, but not every provider uses them, and they shrink the problem rather than remove it. Second, the choice of clock: Horvath, GrimAge and DunedinPACE can assign noticeably different results to the same person — so which one is "your" biological age? Third, the clinical framing is missing: none of these clocks is approved as a diagnostic, there are no reference ranges, no quality standards for consumer labs and no guideline that derives a consequence from the result.

The popular "I'm 50 with the values of a 20-year-old" is therefore almost always marketing poetry: it usually compares individual, favourably chosen readings — not a validated overall picture. A low clock value is a nice signal and not a free pass; a high one is a prompt to check the basics, not a verdict. Basing treatment or supplement decisions on a consumer test is not justified as of today — which is why there is deliberately no product recommendation here.

The good news: the age markers with the best evidence cost nothing and demonstrably respond to training. VO2max is among the strongest known predictors of life expectancy and can be estimated and improved without a lab. Strength (including grip strength), resting heart rate (trackable as a trend under "My metrics" in the app) and waist circumference (a tape measure is enough) plus the classic blood panel from your GP (which metrics are actually worth tracking) together form an age picture that is better validated than any consumer clock.

And "rejuvenation"? The Fitzgerald pilot study on lifestyle and epigenetic age (8 weeks, 43 men, diet/sleep/exercise/relaxation) found a reading a good three years more favourable versus the control group (the change within the program group itself: about two years, statistically only a trend) — an encouraging signal that is no longer alone: the green-Mediterranean DIRECT PLUS trial (2023, randomized, 18 months) and the two-year DAMA trial (2021, randomized, diet plus exercise) independently found slowed methylation clocks too. The cap stays the same: surrogate markers, group-average effects — no forecast for any individual case. What is remarkable is what achieved it: vegetables, protein intake, sleep, movement, stress routines. If you still want to test: treat it as a private long-term experiment with the same provider and the same clock, trend over single value — and health questions still belong in a doctor's office, not in a saliva kit.