This explainer sits alongside the publication's existing piece on what biological age tests measure, found in the explainers hub. That piece covers definitions. This one covers a question readers ask next: if I compare two providers, or take two different test types, why do the numbers disagree, and which one should I trust more?
What are the main families of biological age test?
Three approaches dominate the commercial and research market. Epigenetic clocks measure DNA methylation at selected sites across the genome, usually from a blood or saliva sample, and use a statistical model trained on a reference population to produce an age estimate. Well-known clock models include first-generation clocks trained to predict chronological age and second-generation clocks trained to predict mortality or healthspan outcomes instead.
Blood marker panels combine routine clinical measurements, such as inflammatory markers, lipid values, kidney and liver function markers, and glucose regulation markers, into a composite score. These draw on decades of epidemiological data linking individual markers to disease risk, but the composite score itself is a newer construct and carries its own uncertainty.
Organ-specific panels attempt to estimate the biological age of a single system, commonly the cardiovascular system, the kidneys or the immune system, using a narrower set of markers or imaging proxies. These are the least standardised category, and the published evidence for most organ-age panels is thinner than for whole-body epigenetic clocks.
Each family rests on a different statistical model and a different reference population. A test built on one population's methylation patterns or blood chemistry norms will not necessarily transfer cleanly to another, which matters for UK readers tested against models built largely on other populations. This is a technical point rather than a criticism of any single provider, and it applies across the whole category of consumer biological age testing.
How reproducible are results between repeat samples?
Reproducibility has two distinct meanings here, and conflating them is the most common misunderstanding among readers encountering these tests for the first time. Technical reproducibility asks whether the same sample, run twice through the same assay, gives the same number. Biological reproducibility asks whether two samples taken from the same person on different days give the same number, accounting for normal day-to-day physiological variation.
Published validation work on epigenetic clocks generally reports reasonable technical reproducibility within a single laboratory's pipeline. Reproducibility across different providers using different methylation arrays or different statistical models is less well characterised in independent, published comparisons, and the published literature does not support treating scores from two different commercial providers as interchangeable.
Blood marker composites inherit the natural biological variability of their component markers. A single raised inflammatory marker on the day of sampling, caused by a minor infection or recent exercise, can shift a composite score in a way that has nothing to do with underlying ageing trajectory. This is a known limitation of composite biomarker scores generally, not a flaw unique to any one test. Hydration status, recent meals, sleep the night before and even time of day can all shift individual blood markers enough to move a composite score, independent of any real change in ageing.
What can a result actually tell a reader?
A result can indicate, within the limits of the model used, how a person's measured markers compare with population averages at the time of sampling. For epigenetic clocks validated against mortality in cohort studies, a persistently higher score across repeat testing over months or years has been associated with group-level differences in outcomes in published research.
A result cannot tell an individual reader their personal disease risk with clinical precision, cannot diagnose a condition, and cannot reliably track the effect of a single short intervention, because normal variation and measurement noise can be larger than any plausible short-term biological change. A single test taken once provides limited information beyond curiosity value. Where a reader is tempted to treat a one-off number as a verdict on a diet, a supplement or an exercise programme, it is worth remembering that the model behind the score was built to describe population patterns, not to certify an individual's response to a specific change in behaviour.
How should a reader compare two different test types?
The useful comparison point is not which number is lower, but what each test measures and how it was validated. A test that reports a younger biological age is not automatically more accurate than one that reports an older one, since the two may be built on different reference populations and different statistical assumptions. The table below sets out a decision rule a reader can apply before comparing any two results.
| Test type | What it measures | Validation basis | What a single result supports |
|---|---|---|---|
| Epigenetic clock (first-generation) | DNA methylation pattern correlated with chronological age | Trained to predict age in reference cohorts | A descriptive comparison to population methylation norms, not a risk estimate |
| Epigenetic clock (second-generation) | Methylation pattern trained against mortality or healthspan outcomes | Cohort studies linking score to group-level outcomes | Group-level association evidence only, not individual prognosis |
| Blood marker composite | Combination of routine clinical markers | Epidemiological links between individual markers and disease | A snapshot sensitive to short-term physiological state |
| Organ-specific panel | Markers or imaging proxies for one organ system | Varies widely, often less standardised | Limited, provider-dependent, rarely independently validated |
Why do two providers give different numbers for the same person?
Even within a single family of test, such as two epigenetic clocks, different commercial providers may use different methylation arrays, different panels of sites within the genome, and different proprietary models for converting raw methylation data into an age estimate. None of these choices is disclosed in full detail by every provider, which makes a direct, like-for-like comparison difficult for a reader working from publicly available information alone.
A blood marker panel from one provider and an epigenetic clock from another are measuring different biology entirely, so a discrepancy between the two tells a reader nothing about which is correct. It only tells them that the two tests were built to answer different questions using different data. Treating the lower of two scores as the more reliable one is not supported by how these models are constructed.
What does this not cover?
This explainer does not cover how to interpret a specific commercial provider's report, since methods and proprietary scoring differ between companies and are not always published in full. It does not apply to clinical biomarker testing ordered by a doctor for diagnosis or monitoring of a known condition, which follows different validation standards overseen by clinical regulators rather than consumer testing providers. It is not medical advice, and anyone considering a biological age test for a health decision should discuss it with a doctor rather than acting on a score alone. It also does not cover tests aimed at children, since the reference populations used to build these models are drawn from adult cohorts.
Where to read more
Readers who want the underlying definitions, including what chronological age, biological age and healthspan mean as distinct terms, should see the explainers hub's piece on what biological age tests measure, which this explainer builds on rather than repeats.