Biological age vs chronological age: what the gap means

Chronological age counts time and nothing else: date of birth, today's date, the difference. Biological age is not a measurement but a model estimate: an algorithm reads DNA methylation, blood chemistry or physical performance and returns a number in the same units — years. The gap between those two numbers is the whole point of building the machinery: the calendar runs at the same speed for everyone, the estimates do not.

How the methods themselves work is covered in How biological age is calculated. This article is about the gap: how it is derived, what is known about it, in whom it was shown, and what it does not say.

The gap is a regression residual, not a subtraction

Intuitively, "biological age minus chronological age" looks like plain arithmetic on two numbers. Research more often computes it differently. The clock's estimate is first regressed on chronological age across the whole sample, and the gap is defined as the residual — how far a given person departs from what the model would predict for someone of their calendar age. In the literature this is usually called epigenetic age acceleration.

The distinction is not cosmetic. A plain subtraction carries the model's systematic bias with it: clocks tend to overestimate age in younger participants and underestimate it in older ones, and then "five years younger" is a property of the method rather than of the person. The residual removes the part shared by the sample and leaves the individual deviation.

Hence the first thing worth keeping in mind: the gap is always relative to a sample. It says "compared with people of the same calendar age in this cohort" — and in a different cohort the same person would get a different number.

Epigenetic clocks further distinguish intrinsic from extrinsic acceleration: the first is computed with adjustment for blood cell composition, the second deliberately preserves that component, because age-related shifts in immune cell populations are themselves part of ageing.

What has been shown about the gap, and in whom

The main body of evidence is observational, in adult cohorts. In an analysis across four cohorts (Marioni et al., Genome Biology, 2015; mean participant age in them ranged from 66 to 79 years) higher epigenetic age acceleration was associated with higher mortality, and the association persisted after adjustment for known risk factors — education, social class, hypertension, diabetes, cardiovascular disease, APOE genotype. The same analysis was later extended to thirteen cohorts (Chen et al., Aging, 2016).

Second-generation clocks — PhenoAge (Levine et al., 2018) and GrimAge (Lu et al., 2019) — are built differently. Their target was not calendar age as such but risk: a phenotypic age derived from clinical markers, epigenetic surrogates of plasma proteins, smoking pack-years — with chronological age entering that composite alongside everything else rather than being excluded from it. In cohort data their acceleration tracks outcomes more closely than first-generation clocks do (McCartney et al., Genome Biology, 2021).

All of this describes associations observed in cohorts, not a demonstrated effect. "People with greater acceleration have higher mortality" and "reducing acceleration extends life" are different statements, and the second does not follow from the first.

Moving the gap and gaining clinical benefit are different things

Moving it appears to be possible. In the randomised CALERIE trial (220 participants, two years of caloric restriction) the pace of ageing measured by DunedinPACE shifted, while PhenoAge and GrimAge did not change (Waziry et al., Nature Aging, 2023). In a small open-label study without a control group — nine men, a one-year protocol — several clock estimates went down (Fahy et al., Aging Cell, 2019).

What does not exist is a trial in which shifting the gap led to a change in mortality or morbidity. The CALERIE authors state plainly that further follow-up is required to answer that question. As a surrogate endpoint, epigenetic age acceleration has not been qualified by a regulator.

One person, several clocks, several gaps

This is how the method is built, not a malfunction. The Horvath clock (2013) was trained across many tissues and a wide age range including childhood; the Hannum clock (2013) on the blood of adults aged 19 to 101; PhenoAge targeted a composite clinical measure; DunedinPACE was modelled as a rate of change in a single cohort assessed at ages 26, 32, 38 and 45.

One detail is often lost here. The age estimates themselves correlate strongly across clocks — simply because all of them carry calendar age with them. The accelerations, that is the residuals, correlate only moderately: roughly 0.2 to 0.5 in published comparisons. What diverges is precisely the part the whole exercise is about.

Technical reproducibility adds to this: running two samples from the same person, some clocks return estimates that differ by several years. That is precisely why principal-component versions of the clocks were developed (Higgins-Chen et al., Nature Aging, 2022) — they markedly reduce technical noise in the estimate.

One practical conclusion follows from the whole section: comparing numbers across methods is meaningless. An estimate carries meaning only within one model, one sample type and, where possible, one laboratory.

What the gap does not say

• It is not a diagnosis. Epigenetic age acceleration is not approved by a regulator as a diagnostic test and does not appear in clinical guidelines in that role. Commercial tests are sold as wellness or laboratory-developed tests, which is not the same thing as regulatory approval.
• It is not a prediction for an individual. The associations are established at group level; reviews state directly that on individual-level accuracy these estimates fall short of ordinary clinical measures and should not drive decisions about one person (Apsley et al., Epigenomics, 2025).
• It is not a measure of organ wear. Clocks read methylation in one tissue, most often blood, and the model does not license transferring that result to the body as a whole.
• It is not comparable with someone else's number from another service. See the section above.

A separate note on functional measures, since they enter the same conversation. Gait speed and grip strength are used routinely in geriatrics, as part of sarcopenia and functional status assessment (EWGSOP2, 2019; AWGS, 2019), but they are clinical instruments in their own right, not an approved measure of biological age.

What to do with it in practice

1. Watch the direction, not the absolute number. Same method, same sample type, six to twelve months apart — that is a meaningful comparison. Two different calculators on the same day is not.
2. Keep functional markers alongside it. VO2max, grip strength and gait speed are simple to measure, easy to track yourself, and respond to lifestyle faster than any model estimate.
3. Bring the number to a doctor together with the inputs, not instead of them. The gap is a derived quantity; the conversation is about what it was computed from.

Why a single number cannot settle the question in the first place is covered separately in Biological age: why one number settles nothing.

This article is informational and does not replace consultation with a physician. Decisions about testing and treatment are made with a doctor.