What to Do With a Biological Age Result

You spat in a tube, paid a few hundred dollars, and a website told you your biological age is 47 while your driver’s licence says 52. What should you do?

The honest answer is: almost nothing, and this article explains why in enough detail that you can decide for yourself.

The science underneath these tests is genuine and important. The gap between that science and the number on your report is the whole problem.

What an epigenetic clock actually measures

DNA methylation is a chemical mark attached to DNA that helps determine which genes are active in which cells. The pattern of those marks changes with age in ways that are consistent enough to be modelled.

In 2013, Steve Horvath demonstrated how far this goes. Analyzing 7,844 non-cancer samples from 82 datasets across 51 different tissues and cell types, an elastic net regression selected 353 CpG sites whose methylation predicted chronological age with a correlation of 0.97 and error of 2.9 years in training data, and 0.96 with 3.6 years of error in test data (Horvath, 2013).

That is a genuinely remarkable result. From a blood sample, an algorithm reads your age to within about three and a half years.

The interesting part was never the accuracy, though. It was the errors. When someone’s predicted age ran ahead of their calendar age, that discrepancy — “epigenetic age acceleration” — turned out to carry information about health outcomes. That observation launched the field.

The second generation: pace, not age

Later clocks moved away from predicting chronological age at all.

DunedinPACE was built differently. Rather than training on age, its developers tracked within-individual decline across 19 indicators of organ-system integrity — cardiovascular, metabolic, renal, hepatic, immune, dental, pulmonary — at four time points across two decades in a New Zealand birth cohort, then distilled that trajectory into a single blood test (Belsky et al., 2022).

The result is not an age. It is a rate: how fast your systems are currently declining.

Its technical performance is strong. Test-retest reliability reached an intraclass correlation of 0.96 (95% CI 0.93–0.98) in one dataset and 0.97 (0.94–0.98) in another. In older adults, faster DunedinPACE predicted mortality with hazard ratios of 1.26–1.65 across studies, plus cardiovascular disease, stroke and disability. It added incremental prediction beyond the GrimAge clock.

The authors are also clear about what is missing. They state that establishing clinical utility requires evidence that the measure is modifiable by intervention — and they note their validation cohorts were mostly of white European descent.

“Modifiable by intervention” is precisely the claim the consumer market makes and the literature has not yet established.

The problem with your particular number

Here is where it gets uncomfortable for anyone holding a test result.

A 2025 review in Epigenomics examined whether epigenetic clocks can move from population science to individual clinical use. Its conclusion is blunt: while clocks have proven valuable for population-level research on how behavioral and environmental factors shape disease-risk trajectories, fundamental technical and biological properties of these algorithms prohibit their current use at the individual level (Apsley, Etzel, Ye and Shalev, 2025).

The authors state that clocks fail to meet common standards for clinical utility compared with established biomarkers, and that applying them in individual decision-making can be uninformative and potentially harmful. They go further: even resolving the technical and biological limitations would not, in their view, justify individual-level deployment in the current form.

The reasons split into two groups. Technical: how clocks are constructed, how samples are collected and processed, how data are preprocessed, and how the computation is implemented. Biological: the nature and dynamics of methylation itself, variation across developmental periods, tissue specificity, and sensitivity to environmental and sociodemographic context.

A separate editorial notes that precisely determining biological age and age acceleration remains a significant challenge due to both technical limitations and variability in methylation patterns across populations (Nardini & Di Lena, 2024).

Note the distinction that does the work here. High test-retest reliability — DunedinPACE’s 0.96 — means the assay reproduces itself. It does not mean the number tells you something actionable about your individual future. A bathroom scale accurate to the gram is still not a measure of health.

Why the science is still worth caring about

None of this makes the field a scam. Epigenetic clocks have produced real knowledge: that childhood adversity is associated with faster measured aging, that biological aging rates vary substantially between people of identical chronological age, that this variation predicts morbidity and mortality across large cohorts.

Those are population findings and they are solid. The error is applying a population-calibrated instrument to a single person and treating the output as a personal fact.

What to do with this

• If you have a result, do not act on the number itself. No intervention has been established to change a clock reading in a way that changes your outcome. A result of 47 versus 52 is not a treatment indication.
• If you are considering buying one, ask what you would do differently. If the answer for a “good” result is nothing and for a “bad” result is exercise more, sleep better and check your lipids — you already know the plan. The test is not adding decision value.
• Do not let a favourable result reassure you. This is the concrete harm the Epigenomics authors point at. A young epigenetic age with an ApoB of 130 mg/dL and untreated hypertension is a dangerous piece of comfort.
• Prefer biomarkers with intervention evidence. ApoB, HbA1c, blood pressure and cardiorespiratory fitness all have established thresholds, established treatments, and outcome data behind moving them. Epigenetic clocks have none of the three yet.
• If you test anyway, at least test consistently. Same laboratory, same clock version, same time of day, same season, and ideally more than two data points. Single measurements are the most misleading form of this data.
• Watch the field for one specific development: randomized trials showing that an intervention changes a clock reading and that the change tracks a hard outcome. Until that exists, these are research instruments sold as personal ones.

The most useful thing a biological age result can do for you is prompt the question of what you are actually doing about your health. That is a real benefit. It is also available for free.