Two people buy a biological age test in the same week. One gets back a number derived from chemical tags on their DNA. The other gets a number derived from sugar structures on their antibodies. Both are sold as "biological age," both cost roughly the same, and the two numbers can disagree by a decade without either being wrong.
They are measuring different biology.
The Verdict
The two families, side by side
| Methylation (epigenetic) clocks | Glycan / immune age | Protein-based immune panels | |
|---|---|---|---|
| What it reads | DNA methylation at specific CpG sites in white blood cells | Sugar structures (glycans) attached to circulating IgG antibodies | Panels of circulating immune and inflammatory proteins |
| What it is trained to predict | Chronological age, or mortality and disease risk depending on the clock | Inflammatory ageing of the immune system | Inflammatory burden and immune decline |
| Typical sample | Blood spot or blood draw | Finger-prick blood spot | Blood draw |
| Speed of response to intervention | Slow. Most second-generation clocks move over many months. | Faster, though the figures vary: providers cite 6–12 months for lifestyle change and shorter windows for hormonal or drug interventions. | Variable, and confounded by any acute illness |
| Test-retest noise | Meaningful. Single-CpG noise is a known problem; principal-component versions reduce it. | Reported as lower, though independently published reliability data is thinner. | High for single markers such as CRP |
| Independent validation | Extensive. Thousands of published papers across large cohorts. | Real but narrower, and heavily weighted toward the group commercialising it. | Strong for individual markers, weaker as a composite age score |
Not all methylation clocks are the same test
Treating "epigenetic clock" as one thing causes most of the confusion in this category. The generation of the clock determines what the number means.
| Clock | Generation | Trained to predict | What that means in practice |
|---|---|---|---|
| Horvath (2013) | First generation | Chronological age across many tissues | Accurate at guessing age. That is the problem: a clock trained to match your birthday cannot easily tell you that you are ageing badly. |
| Hannum (2013) | First generation | Chronological age in blood | Same limitation. |
| PhenoAge (2018) | Second generation | Clinical measures of ageing, then mortality | Predicts mortality and disease better than first-generation clocks. |
| GrimAge (2019) | Second generation | Time to death, plus smoking-pack-years proxies | Currently the strongest mortality prediction among widely used clocks. |
| DunedinPACE (2022) | Pace-of-ageing | Rate of decline across 19 biomarkers tracked over decades | Reports a speed rather than an age. The most sensible design for tracking an intervention. |
The first-generation clocks were trained to guess chronological age, and they do it accurately. That accuracy is a design limitation once you want a health signal. A clock optimised to output your calendar age is optimised to ignore the ways you differ from it.
Second-generation clocks fixed this by training against mortality and clinical measures instead. DunedinPACE, developed from a birth cohort followed for decades, went further and reports a rate rather than an age. A pace of 1.0 means you are ageing at one biological year per calendar year. For anyone trying to detect whether an intervention changed anything, a rate is the more sensible output.
If a provider will not tell you which clock generates your number, that is the answer to whether the number is worth acting on.
The response ladder: what each layer can resolve
Response speed is easier to reason about as a ladder than as a two-way comparison. Standard blood markers sit at the fast end, methylation clocks at the slow end, and glycan and immune measures in between. Where a test sits determines what it can resolve and what it will mistake for progress.
| Layer | Response window | What it is reading | What it cannot separate out |
|---|---|---|---|
| Standard blood markers (hsCRP, fasting insulin, HbA1c, lipids) | Weeks to a few months | Current physiological state | Almost none of it is specific to ageing. A chest infection moves hsCRP more than a year of ageing does |
| Glycan and immune-age measures | Roughly 6 to 12 months for lifestyle change, per provider data | Structural change to circulating antibodies, which accumulates over months | Still sensitive to recent illness, autoimmune flares and heavy training blocks |
| Methylation clocks | Many months to years | Slow, integrated change in DNA methylation patterns | Test-retest noise sets a floor. Below that floor a change cannot be distinguished from the assay |
The middle rung is the one people buy for the wrong reason. A glycan measure can plausibly resolve a six-month intervention that a methylation clock is still too noisy to detect, which is a genuine capability. It will also register recovery from a chest infection as improvement, because the molecule it reads responds to both. Catching a change and knowing what caused it are different problems, and a faster test only helps with the first.
Hormonal interventions sit at the fast end of that middle rung, resolving sooner than diet or exercise because sex hormones act directly on IgG glycosylation. That makes them the clearest case where a glycan test catches a shift a methylation clock is still too slow to see, and equally the case where a moving score tells you something changed rather than that a therapy is working. The FAQ below covers how those decisions should actually be made.
Reading down the ladder gives the practical rule. If you want to know whether something you changed is having any effect at all within a season, the cheapest answer is on a standard panel. If you want to know whether your rate of ageing has shifted, that question is answered in years, on one assay, at one lab.
What each family is genuinely good for
- Risk stratification: a second-generation methylation clock. GrimAge and PhenoAge have been validated against mortality in large independent cohorts, which no glycan measure currently matches.
- Tracking an intervention over 3 to 6 months: a glycan or immune measure, with the caveat above, or DunedinPACE if you accept a longer interval.
- Understanding which body systems are ageing unevenly: organ-specific methylation panels, which are newer and less validated than the whole-body clocks but answer a question the others cannot.
- Making a clinical decision: neither. That still runs through ApoB, Lp(a), blood pressure, HbA1c and fasting insulin, all of which have thresholds attached to actual guidelines.
The question to ask before buying either
Ask what the test's test-retest reliability is, in the provider's own hands, on split samples from one blood draw. This single question separates the providers who have measured their own noise from those who have not.
It matters because of how these results get used. Someone tests, changes their training and diet for four months, retests, and sees their biological age drop by two years. If the assay's own variation between duplicate samples is a year and a half, most of that movement is measurement noise wearing the costume of progress. Providers that publish reliability figures, and those that have moved to principal-component clock versions to reduce noise, are the ones taking the problem seriously.
Our consistent position across this category is that a single biological age result is close to uninformative, and a trend of three or more results on the same assay, at the same lab, at the same time of year, is where any signal lives. That is also the framing that makes the cost obvious: the meaningful purchase is a multi-year subscription, not a one-off test.
What neither family can tell you
- Whether you have a specific disease. These are population-derived summary measures, not diagnostics.
- Which intervention caused a change, since a single-arm experiment on yourself has no control group.
- How long you will live. The mortality associations are real at population scale and weak at individual scale.
- Whether a supplement worked. No NAD+ precursor or senolytic has been shown to move a validated clock in a controlled trial.
When to take a result to a clinician
- A pace of ageing above 1.0 that persists across repeat tests, which is the most interpretable adverse output these tests produce and worth raising alongside your conventional risk markers.
- Any result arriving with a genuinely abnormal conventional marker, since that marker, not the composite age, is what a clinician can act on.
- An organ-specific score flagged repeatedly, which warrants checking the established marker for that system rather than acting on the score itself.
- Any symptom you were planning to mention anyway. A biological age result is not a substitute for describing what you are experiencing.
Where to read the research
- Belsky et al. 2022, eLife: DunedinPACE, the pace-of-ageing measure.
- Levine et al. 2018, Aging: PhenoAge.
- Lu et al. 2019, Aging: GrimAge.
- Horvath 2013, Genome Biology: the first multi-tissue clock.
- Higgins-Chen et al. 2022, Nature Aging: the principal-component clocks built to reduce test-retest noise.
Frequently Asked Questions
Why do glycans respond faster than a methylation clock but slower than a blood test?
Because each layer sits at a different point on a response ladder, and position on that ladder is set by how quickly the underlying molecule turns over. Standard inflammatory and metabolic markers reflect current state and move in weeks. IgG glycan structures accumulate over months, so they lag current state but resolve a change faster than methylation does. Methylation patterns integrate over years by design. The ordering is not a quality ranking: each layer trades specificity to ageing against speed of response, and no test on the market escapes that trade.
Can a glycan test detect a change an epigenetic clock would miss?
Within a short intervention window, yes, and that is the strongest case for buying one. If you change something for six months, a glycan measure can plausibly resolve the shift while a methylation clock is still inside its own noise floor. The reverse is also true and less often stated: a glycan test can register a change that has nothing to do with ageing, such as recovery from an infection, and report it as improvement. Detecting a change and attributing it correctly are separate problems.
How quickly can these tests register a hormonal intervention?
Faster than a lifestyle change, on the immune-age side. Providers of glycan tests report shorter response windows for hormonal interventions such as hormone therapy or testosterone therapy than for diet and exercise, which is consistent with sex hormones having direct effects on IgG glycosylation. Independent replication of those windows is thin, and none of it establishes that a shift in the score means the therapy is working. Decisions about starting, adjusting or stopping hormone therapy run through symptoms and standard markers with a prescribing clinician, not through a biological age score.
Why is it hard to act on an epigenetic clock result after an intervention?
Three problems stack. The biology moves slowly, so a three-month gap between draws is too short for a real signal to appear. The assay noise on some clocks is large enough that a change of a year or more can be measurement variation. And a single-arm experiment on yourself has no control group, so even a real change cannot be attributed to the thing you changed. The practical answer is to retest annually on the same assay and read the trend across three or more results.
What is the difference between a methylation clock and an immune age test?
They read different molecules and answer different questions. A methylation clock measures chemical tags on DNA in white blood cells and estimates either your age or your mortality risk, depending on which clock is used. An immune age test, including glycan-based tests, measures markers of inflammatory ageing in the immune system. A methylation clock is better validated against hard outcomes such as mortality. Immune and glycan measures are reported to respond faster to lifestyle change.
Which biological age test responds fastest to lifestyle changes?
Glycan and immune-based tests are marketed on faster response, and the intervention data providers publish does suggest they move within 6 to 12 months for a lifestyle change, while most methylation clocks move more slowly than that. Shorter windows quoted in marketing usually refer to hormonal or drug interventions rather than diet and exercise. That advantage comes with a trade-off. A marker that moves quickly is also more easily moved by things that are not ageing, including a recent infection, an injury, or a short-term change in inflammation. Speed of response and specificity to ageing pull in opposite directions.
Is a methylation clock accurate enough to retest every few months?
Usually not. Two problems compound. Test-retest variation on some methylation clocks is large enough that a repeat sample from the same blood draw can differ by a year or more, and the underlying biology genuinely moves slowly. Principal-component versions of these clocks were developed specifically to cut that noise. If a provider offers quarterly retesting on a standard clock, ask what their measured test-retest reliability is before treating a change as real.
Does a glycan test measure the same thing as the inflammatory markers on a blood panel?
No. Standard inflammatory markers on a longevity panel, such as hsCRP, measure current inflammatory activity and swing widely with infection or injury. IgG glycan profiling measures structural changes to antibodies that develop over longer periods. They are related but not interchangeable, and a normal hsCRP does not tell you what a glycan profile would show.
Which test should I buy?
For most people, a second-generation methylation clock such as GrimAge, or a pace measure such as DunedinPACE. Those carry the deepest independent validation against mortality and disease risk, which is the question most buyers are actually asking. A glycan or immune measure is the narrower purchase, justified when you specifically want to track inflammatory ageing across a defined intervention and are prepared to control for recent illness and training load. Buying both mostly buys two numbers that disagree.
Do these tests change what a doctor would do?
Rarely on their own. No major clinical guideline incorporates biological age testing into screening or treatment decisions. The markers with established clinical thresholds remain the standard ones: ApoB, Lp(a), blood pressure, HbA1c, and fasting insulin. A biological age result is best understood as a research-grade summary measure that may motivate behaviour change, not as a test that redirects medical care.