Testing & Measurement · Updated September 2026
Epigenetic clocks are real science with a consumer-product problem: several published analyses have found that repeat measurements on the same sample can differ by years. Here is what each test measures and how much weight it can carry.
There is no single biological age. There are several distinct measurements that each correlate with age-related health outcomes to different degrees, and consumer products present all of them with the same confident single-number output.
DNA methylation clocks are the technology behind most premium consumer tests. Methylation is a chemical modification to DNA that changes in patterned ways across the lifespan, and an algorithm trained on those patterns estimates age. First-generation clocks — Horvath and Hannum — were trained to predict chronological age, which means a perfect result is one that tells you what your birth certificate already says. Their value comes from the residual: whether you read older or younger than your actual age.
Second-generation clocks — PhenoAge and GrimAge — were trained differently, on clinical biomarkers and mortality outcomes rather than on chronological age. That is a meaningful improvement, and these clocks consistently predict mortality and disease risk better than the first generation does. A third generation, including pace-of-ageing measures such as DunedinPACE, estimates the rate of ageing rather than a point value.
Phenotypic age is computed from nine routine blood markers plus chronological age, using a published equation. It requires no sequencing, can often be derived from a blood panel you already have, and correlates with outcomes surprisingly well for something so cheap.
Telomere length is the oldest consumer offering in this category and the weakest. Telomeres shorten with cell division and telomere length does correlate with age at the population level, but measurement variability is high and individual-level predictive value is poor. It is largely a legacy product.
Several published analyses have examined the technical reliability of epigenetic clocks by running the same sample multiple times. The results have been sobering: first-generation clocks in particular can produce estimates differing by several years on identical input, driven by measurement noise in the underlying methylation arrays.
If a repeat measurement of the same blood can move by several years, then a three-year improvement after six months of a new supplement regimen is not distinguishable from noise. This is the single most important thing to understand before paying for one of these tests, and it is the thing least likely to appear in the marketing.
Newer processing methods and principal-component-based versions of the clocks substantially improve reliability, and providers that use them are meaningfully better. Ask which clock version and which processing pipeline a test uses — if the answer is not available, treat the number as entertainment.
| Test type | Measures | Predicts outcomes? | Reliability | Cost band |
|---|---|---|---|---|
| 1st-gen methylation clock | Chronological age proxy | Weakly | Poor without PC methods | $$$ |
| 2nd-gen clock (PhenoAge, GrimAge) | Mortality-trained composite | Well | Better | $$$ |
| Pace-of-ageing (DunedinPACE) | Rate of ageing | Well | Better | $$$ |
| Phenotypic age from blood | 9 clinical markers + age | Well | Good | $ |
| Telomere length | Average telomere length | Poorly at individual level | Poor | $$ |
| Standard blood panel | Actual clinical risk markers | Well, and actionably | Good | $ |
The honest case for buying an epigenetic test is curiosity plus a willingness to treat the number as a rough, noisy signal rather than a score. If you have already optimised the things known to move health outcomes — training, sleep, diet, blood pressure, lipids, not smoking — and you want another data point, it is a defensible purchase.
The case against is that almost every actionable input to your biological age is already visible in a standard blood panel that costs a fraction as much, and that the interventions supported by evidence are the same ones regardless of what the clock says. A methylation result showing you are three years older than your chronological age does not tell you what to do differently; a fasting glucose of 108 and an ApoB of 130 do.
If you are going to track biological age over time, the procedural rules matter more than the brand. Use the same provider, the same clock version and the same sample type every time — results are not comparable across companies, and a saliva result cannot be compared to a blood result. Sample at the same time of day, and avoid testing during or shortly after an acute illness, since inflammation shifts several of these measures.
Retest no more often than annually. Given known measurement variability, quarterly testing will show you noise and invite you to attribute it to whatever you changed that quarter, which is how expensive supplement habits get built on nothing.
Second-generation clocks such as PhenoAge and GrimAge, or a pace-of-ageing measure, rather than a first-generation chronological-age clock.
Principal-component or otherwise noise-reduced processing, which substantially improves test-retest reliability over standard array processing.
Analysis performed by an accredited laboratory, with the lab named rather than implied.
Whether you can export the underlying data rather than only a marketing dashboard number, which matters if you want a second opinion later.
Ordered by how much the result can actually inform a decision rather than by price. The cheapest option sits near the top, which is unusual in this category and worth stating plainly.
Phenotypic age is computed from routine markers — albumin, creatinine, glucose, CRP, lymphocyte percentage, mean cell volume, red cell distribution width, alkaline phosphatase and white cell count — plus your chronological age. It predicts outcomes well, costs a fraction of a methylation test, and every input is independently actionable in a way a clock output is not.
View on AmazonThe premium consumer option. Worth buying only from a provider that names its clock version and uses noise-reduced processing, and worth reading as a broad signal rather than a precise number. Repeat annually, never quarterly.
View on AmazonHigh-sensitivity C-reactive protein is one of the few inflammatory markers with an established role in cardiovascular risk assessment, and it feeds several biological age composites. Cheap, widely available, and interpretable on its own.
View on AmazonApoB counts atherogenic lipoprotein particles directly rather than the cholesterol they carry, and is a strong predictor of atherosclerotic risk — in some analyses better than LDL-C where the two disagree. It has been historically under-ordered relative to its value.
View on AmazonGrip strength is one of the better-validated functional predictors of all-cause mortality and disability in older adults, it costs almost nothing to measure, and unlike a methylation clock it responds visibly to training. As a biological age proxy it is cheap, repeatable and directly modifiable.
View on AmazonThe underlying science is real; the consumer implementation is noisier than the marketing suggests. Published reliability analyses have found that first-generation epigenetic clocks can produce estimates differing by several years when the same sample is measured repeatedly. Second-generation clocks trained on health outcomes, and newer principal-component processing methods, are meaningfully more reliable — which is why the clock version and processing pipeline matter more than the brand.
First-generation clocks such as Horvath and Hannum were trained to predict chronological age, so their output is most useful as a residual — whether you read older or younger than your actual age. Second-generation clocks such as PhenoAge and GrimAge were trained on clinical biomarkers and mortality outcomes instead, and they predict disease risk and mortality considerably better.
Generally no. Telomere length correlates with age at the population level, but measurement variability is high and individual-level predictive value is poor. It is the oldest consumer offering in this category and largely superseded by methylation clocks and by phenotypic age computed from routine blood markers.
Annually at most. Given the known measurement variability, testing quarterly will mostly show you noise, and the temptation to attribute that noise to whatever you changed that quarter is exactly how expensive supplement habits get built on nothing. Standard blood markers can be retested more frequently because they are more reliable and directly actionable.
The interventions with evidence behind them are the unglamorous ones: regular exercise including resistance training, adequate sleep, not smoking, managing blood pressure and lipids, and maintaining a healthy body composition. Studies have reported changes in clock estimates alongside lifestyle interventions, but given measurement reliability limits, small reported reductions should be treated cautiously. No supplement has been shown to reliably lower biological age in humans.
Take the 7-question quiz to get your personalized supplement and lifestyle stack — tailored to your age, fitness level, and longevity concerns.
Take the QuizA 28-day fillable sleep log built around the seven levers that actually move sleep, so you test one per week and see what it did.