"Fitness age" on a wearable is not a biological-age test

TL;DR

• A fitness age is one number in disguise: an estimated cardiorespiratory fitness value. Everything else the device shows you is an input to that estimate, not a second source of evidence.
• The "age" is produced by a lookup, not a measurement - the algorithm finds the chronological age at which the average person of your sex has your estimated fitness and reports that back, which makes it a percentile with a years label: a statement about a reference population, not about the state of your tissue.
• Its underlying quantity is genuinely well validated. Cardiorespiratory fitness is among the strongest predictors of mortality in the epidemiological literature. The estimate of it on your wrist is much softer than the quantity itself.
• Epigenetic and organ clocks measure something categorically different - molecular state, fitted against different targets - which is why the two can disagree without either being wrong.
• The years unit amplifies error. Population median fitness declines slowly per year of age, so a small estimation error maps onto a large-looking number of years.
• Use it as a trend on unchanged hardware and habits, not as a level - and do not treat it as a health verdict in either direction.
• Training is the lever, and some people also ask about medical options. A new section sets out recovery, sleep, GLP-1 and growth-hormone-axis options with their evidence and status, as conversations for a physician rather than ways to move the number.

Educational content, not medical advice. No doses or protocols appear. No device, brand, model or product is named, compared or recommended anywhere on this page, and nothing here is an accuracy claim about any specific product.

---

The experience is now common enough to be a genre. You are forty-seven. Your watch says your fitness age is thirty-four. It is a pleasant thing to be told, and the obvious next question is whether it means anything - specifically, whether it is the same kind of claim as the biological-age tests you have read about, and whether the thirteen-year gap is a finding.

The short answer is that it is a different kind of claim entirely, and the difference is not a matter of degree. A wearable fitness age is a fitness estimate expressed in unusual units. It is not a measurement of ageing, it was never designed to be one, and the companies computing it have generally been more careful about saying so than the people quoting it.

This page owns that one question: what the consumer-device "age" score is computed from, and why that construction means it is not an ageing measurement. Two adjacent questions live elsewhere in the library and are not re-argued here. Why two biological-age tests give you two different answers - training targets, reference distributions, assay noise - is worked through in why biological age tests disagree. What the families of biological-age test are actually built from is covered in what biological age tests measure.

What a wearable "age" score is computed from

Strip away the presentation and there are three input categories. Only the first one carries the age number.

Estimated VO₂max

Cardiorespiratory fitness is conventionally expressed as maximal oxygen uptake - the volume of oxygen consumed per unit of body mass per unit of time at maximal effort. The reference method for obtaining it is a graded exercise test to volitional exhaustion with expired-gas analysis, in a laboratory, with a mask on your face.

A wrist device does none of that. It estimates the same quantity from a statistical model, typically combining:

• heart rate during submaximal activity, sampled optically at the wrist;
• external workload inferred from satellite-derived pace, or from accelerometer-derived cadence and motion when pace is unavailable;
• demographic and anthropometric inputs you typed in - age, sex, height, weight, sometimes a self-reported activity level;
• an assumed maximal heart rate, usually derived from an age-based formula rather than measured.

The logic is the relationship between heart rate and oxygen uptake: for a given external workload, a fitter person operates at a lower fraction of their maximum. A model fitted to that relationship in a development population can return a plausible VO₂max estimate without anyone reaching maximum.

Where this goes soft is worth being specific about, because it is the whole basis of the number:

• The maximal heart rate is usually assumed, not known. Age-based prediction equations carry a wide standard error at the individual level, and an error in the assumed maximum propagates directly into the fitness estimate.
• Optical heart-rate sensing at the wrist degrades under conditions that matter. Accuracy falls with motion artefact, at higher intensities, in cold, and - as Bent and colleagues showed across devices - differentially by skin tone and other physiological characteristics. This is a property of the sensing modality, not of any one product.
• Validation studies typically report reasonable agreement with laboratory testing at group level and much wider limits of agreement at individual level. The pattern is regression toward the development population's mean: over-estimation in less-fit individuals, under-estimation in highly fit ones. Individual errors are frequently large enough to move a person a whole fitness category.
• The estimate needs a specific kind of data to be any good at all. Sustained, steady, outdoor, pace-measurable aerobic effort is what these models were built on. A user whose training is strength work, intervals, swimming or classes may be handed a confident number generated from very little usable signal.

Resting heart rate and heart-rate variability

These usually feed the model, and they feed the other scores on the device - the recovery and readiness numbers - rather than the age.

Resting heart rate is the more solid of the two: reasonably reproducible when measured under consistent conditions, and inversely associated with fitness. It also moves with illness, alcohol, heat, altitude, dehydration, stress, caffeine and medication, which is why a single value is uninformative and a sustained shift is worth noticing.

Heart-rate variability is far more fragile than its reputation. Its value depends on which metric is computed, over what window, at what time of day, and - heavily - on respiration rate, posture and measurement method. Within-person day-to-day variation is large. Values from two different measurement approaches are not interchangeable, and there is no meaningful population "normal" to compare yourself against. HRV is legitimately useful as a personal baseline tracked with unchanged method; it is not a biomarker with an external reference range, and it is not an ageing measurement.

Readiness, recovery and energy-reserve composites - whatever a given manufacturer calls them - deserve one honest sentence. They are proprietary weighted combinations of resting heart rate, heart-rate variability, sleep estimates and recent training load. The weights are undisclosed, they differ between manufacturers, they change with firmware updates, and they have not been validated against any health outcome. They are behavioural nudges. Treating one as a physiological measurement is a category error.

Activity volume

Step counts, active minutes, intensity minutes and weekly training load contribute to some fitness-age implementations, either directly or by determining whether the device has enough data to update its estimate. Activity volume is the input with the best claim to external validity - self-reported and device-measured physical activity both predict outcomes in cohort studies - but as a component of an age score it is doing something much more modest than it appears: it is helping to place you on a fitness distribution.

Why a fitness estimate produces an age number: the population-percentile trick

Here is the step that explains everything else, and it is not hidden - it is just rarely explained.

The device has an estimated VO₂max. It also has a normative reference table: population average or median fitness by sex and age band, drawn from a research cohort. The concept and the method were developed in the exercise-science literature, notably in work from the Norwegian HUNT cohort, where a "fitness age" was proposed precisely as an accessible way to communicate a percentile to non-specialists.

The computation is then a lookup in reverse. The algorithm asks: at what chronological age does the reference population's average fitness equal this person's estimated fitness? That age is returned as your fitness age.

Four consequences follow directly, and they are all structural rather than fixable:

It is a comparison, not a measurement. "Fitness age thirty-four" means: your estimated fitness resembles that of an average thirty-four-year-old of your sex in some reference population. It contains no information about your cells, your methylation, your organ function or your rate of ageing. It is a percentile wearing a different coat.

The number depends on a population you did not choose and cannot see. Change the reference cohort - different country, different era, different recruitment - and your fitness age changes without anything about you changing. A fitter reference population makes you older. A less fit one makes you younger. This is the same undisclosed-reference-distribution problem that makes "years younger" claims slippery in the biological-age market generally; the mechanics of that are set out in why biological age tests disagree.

The scale is bounded and compressed at the ends. Implementations floor and cap the output, because the reference table runs out. A very fit forty-seven-year-old can hit the floor and stay there regardless of further improvement; the number then stops carrying information precisely for the people most motivated by it.

There is only one real input. Because the age is a transformation of a single estimated quantity, it adds no information to that quantity. If you know your estimated VO₂max and its trend, the age label tells you nothing further. It is strictly a presentation layer.

Worth saying clearly, because this page is not an argument that the number is worthless: the quantity underneath it is one of the best-validated prognostic markers in medicine. Cardiorespiratory fitness predicts all-cause and cardiovascular mortality strongly and consistently, across cohorts, continents and decades of follow-up - better than most of what is sold as a biological-age test. That evidence, and the important gap between predicting an outcome and changing it by moving the marker, is the subject of functional markers of biological age. The problem is not the biology. The problem is two layers of translation: a laboratory quantity replaced by a wrist estimate, and an estimate relabelled as an age.

What it is genuinely useful for: trend, not level

Most of the useful signal survives if you demote the number from verdict to feedback.

A trend in your own estimate, on unchanged hardware and unchanged usage, is real information. Systematic error that is stable across measurements largely cancels when you look at direction. If your estimate has moved consistently in one direction across several months while you kept wearing the same device the same way for the same kinds of activity, something has probably actually changed.

It is good adherence feedback. The number responding to training is a reinforcement mechanism, and reinforcement mechanisms that keep people training are not trivial. That is a behavioural benefit and it is worth having.

A sustained shift in resting heart rate is worth attention. Not a single high morning - a persistent change in the baseline, which can reflect illness, overreaching, alcohol pattern, medication, or something that warrants a clinician's opinion.

Large changes are detectable; small ones are not. A previously sedentary person who begins training will see a genuine move. A well-trained person hunting a two-year improvement is reading noise.

Three uses that are not supported by the construction of the number: comparing your score with someone on a different device (different algorithms, different reference tables, incomparable outputs); reading the level as a health status; and treating it as evidence about your rate of ageing.

What an epigenetic or organ clock measures instead

The contrast is what makes the two incomparable, so it is worth stating properly rather than gesturing at.

Epigenetic clocks are statistical models built on DNA methylation at selected CpG sites. The generations differ in what they were fitted against: first-generation clocks against chronological age; second-generation clocks against mortality or clinical-biomarker composites; third-generation approaches against the rate of longitudinal physiological decline rather than a level. The output is expressed in years because years are intuitive - the unit is a presentation choice there too, which is one of the few genuine similarities with a fitness age.

Organ clocks are newer, generally proteomic, and report separate ages for separate organ systems on the basis that organs do not age in step.

What both share, and what a fitness age does not have: they are attempts to read a molecular state that is not otherwise visible, fitted against an ageing-relevant target. What neither has, and what fitness has: direct, repeatedly replicated, multi-decade outcome validation of the underlying measurement, and straightforward modifiability.

Neither category is a validated surrogate endpoint. No clock is accepted by regulators as a stand-in for healthspan or lifespan, and improving any of these numbers has not been shown to change an outcome. The full treatment of what each family is built from, and what a result can and cannot support, is in what biological age tests measure.

Where the two can disagree completely, and why neither is lying

A well-trained forty-seven-year-old can carry a flattering fitness age and an unflattering epigenetic result. The reverse happens too. Both can be internally valid at the same time, for reasons that follow from the constructions above.

They are looking at different things. A fitness age reports one physiological domain - oxygen transport and utilisation capacity - via an estimate. A methylation clock reports a molecular pattern influenced by inflammatory burden, adiposity, smoking history, glycaemic exposure, genetics and much else. High aerobic capacity does not oblige the molecular layer to look favourable, and a favourable molecular layer does not oblige capacity to be high.

They were fitted against different targets. One is anchored to a normative fitness distribution; the other to chronological age, a mortality composite, or a rate of decline. Two models with different targets, different inputs and different reference populations producing different numbers in the same unit is arithmetic, not contradiction.

Their error structures are unrelated. A wearable's error is dominated by heart-rate sensing, assumed maximal heart rate and the quality of the activity data. A clock's error is dominated by assay reproducibility and genuine day-to-day biological movement. These are independent, so the two results can diverge simply because both are noisy in different ways.

One is modifiable on a timescale you can observe; the other largely is not. Training moves measured fitness over weeks to months. Whether anything reliably moves a methylation clock, and by how much, is a considerably more contested question - and it is a question about a different kind of object.

The practical instruction that falls out of this is the same one that falls out of the disagreement literature: do not average them, do not pick the flattering one, and do not treat either as a verdict. Averaging is specifically wrong, because the mean of two incomparable quantities has no interpretation.

The measurement-error problem: how much of a change is real?

This is the question that decides whether the number is usable, and the honest answer is that most of the small movements people react to are not real.

Four sources of movement, only one of which is you:

Estimation error in the algorithm. Wrist-derived VO₂max estimates carry individual-level error large enough, in validation work, to shift someone across a fitness category. That error does not vanish on re-measurement; it partly re-rolls.

Genuine biological day-to-day variation. Resting heart rate, heart-rate variability and submaximal heart-rate response all move with sleep, hydration, heat, alcohol, caffeine, illness and stress. On any given day the inputs are different even if your fitness is not.

Input drift. A device worn tighter or looser, a change in the mix of activities, a corrected weight entry, a switch to a different wrist, a change of season and therefore ambient temperature - all of these move the number. And firmware and algorithm updates silently change the computation, which means a year-long trend may not be measuring one consistent quantity at all. This is the least discussed and probably the most consequential source of spurious change.

The years unit amplifies everything above. This is the crux. Population average fitness declines gradually with each year of adult age - a fraction of a fitness unit per year. So the reference table converts a small change in estimated fitness into a large-looking change in years. A modest estimation error becomes "five years younger." The dramatic-sounding output is a property of the unit conversion, not of the size of the underlying change.

Putting those together gives a defensible working rule: treat direction sustained over months, on unchanged hardware and unchanged usage, as possibly real; treat anything smaller as noise. And if you want to know whether your fitness actually changed, the answer is not available from a re-read of the score - it comes from a measured test, which is a different undertaking.

What to do with the number you already have

Practical, and deliberately unexciting.

Stop reading the age and read the inputs. Your estimated VO₂max trend, your resting heart rate trend and your weekly activity volume are the substance. The age label adds nothing to them.

Ask whether the device has usable data about you. If your training is mostly strength work, intervals, swimming or classes, the fitness estimate was built from very little of the signal it needs, and the number deserves proportionally less weight.

If a flattering number made you comfortable, notice what it does not cover. Aerobic capacity is one domain. Blood pressure, lipids including the apolipoprotein measures, glycaemic regulation, body composition and strength are separate domains, and a low fitness age does not speak to any of them. Glucose regulation in particular has its own well-worn consumer-measurement trap - genuinely informative, easy to over-read - covered in CGMs for people without diabetes.

If an unflattering number worried you, escalate to real measurement rather than a second gadget. Blood pressure measured properly, standard bloodwork, and where clinically indicated a supervised exercise test, will tell you far more than any wrist estimate. Buying a second device to arbitrate between two estimates produces two estimates.

If you want a functional baseline you can trust, the well-validated versions are cheap. Grip strength, gait speed and a chair-stand test are validated against mortality and disability directly, need no algorithm, and are covered in functional markers of biological age.

Interpret anything you find with a clinician. This is not boilerplate. The failure mode here is not a wrong number; it is a person making decisions about their health from a percentile they have mistaken for a diagnosis.

Where peptides fit: options to discuss with a physician

Nothing here changes the argument above. A fitness age is a percentile expressed in years, and no compound or medicine is a way to "lower" it. The things underneath it are real, though: aerobic capacity, muscle, and recovering well enough between sessions to keep training. Some people pursuing those goals ask about peptides and related medicines alongside training, sleep and nutrition, so the options are set out below with the evidence tier and regulatory status stated plainly. Each is a conversation to have with a physician who knows your history, not something to start on your own. Vivre Labs offers access, and we recommend beginning with baseline bloodwork (what to test); how we grade evidence is explained in peptide evidence tiers.

Recovery between sessions

• BPC-157 is a synthetic peptide studied mainly for tendon, ligament and muscle injury in rats. Evidence: animal studies, plus one uncontrolled retrospective chart review of knee pain in which 14 of 16 people surveyed by phone reported improvement, with subjective outcomes only and no control group. Status: investigational, not FDA-approved.
• TB-500 is a synthetic peptide related to thymosin beta-4. Evidence: animal and cell studies of tissue repair; no controlled human trial of recovery. Status: investigational, not FDA-approved.
• GLOW is a protocol combining BPC-157, TB-500 and GHK-Cu, aimed at recovery and skin. Its evidence is that of its components, which is mostly animal work; injectable GHK-Cu is also investigational.

In July 2026 an FDA advisory committee recommended BPC-157 and TB-500 for the 503A compounding list. The vote is advisory only: neither has been added, and an interim enforcement-discretion policy sent to OMB on 14 October 2026 has not been issued. More in recovery peptides.

Sleep

Sleep is where much of recovery happens, and it also feeds the resting heart rate and heart-rate variability numbers your device reports.

• DSIP (delta sleep-inducing peptide). Evidence: human sleep studies that are small, old and inconsistent. Status: investigational; the same FDA advisory committee voted 6 to 7 against recommending it for the 503A list.
• Epithalon is a short synthetic peptide studied for sleep and circadian rhythm. Evidence: mostly small Russian studies and animal work. Status: investigational, not FDA-approved.
• The lean mass and sleep quality protocol pairs CJC-1295 with ipamorelin, two compounds acting on the growth-hormone axis described below. Both are investigational.

GLP-1 medicines, for people with obesity

Semaglutide and tirzepatide are FDA-approved for type 2 diabetes and for chronic weight management in eligible adults, with some further specific indications. Their effect on body weight and metabolic markers is established in large randomised trials. Two points tie them to this article. First, VO₂max is expressed per unit of body mass, so losing weight raises the estimate arithmetically even if the heart and muscles have not changed; a better fitness age after weight loss is partly a unit effect. Second, part of the weight lost on these medicines is lean mass, and preserving muscle depends on resistance training and adequate protein, which is a plan to make with your physician (GLP-1s, resistance training and protein). They are prescription medicines for defined indications, not fitness aids.

The growth-hormone axis (GHRH and related peptides)

These compounds act on the body's own growth-hormone release, which bears on muscle, body composition and deep sleep; the background is in the GH/IGF-1 axis.

• CJC-1295 is a synthetic analogue of growth hormone-releasing hormone (GHRH), and ipamorelin is a ghrelin-receptor agonist that prompts growth-hormone release. Evidence: small human pharmacology studies show compounds in these classes raise growth hormone and IGF-1; no controlled trial shows better fitness, muscle or sleep outcomes in healthy adults. Status: both investigational, not FDA-approved.
• Sermorelin is an older GHRH analogue. An FDA-approved sermorelin product existed and was discontinued in 2008 for reasons unrelated to safety; no approved product is currently marketed.
• Tesamorelin is a GHRH analogue FDA-approved to reduce excess abdominal fat in people with HIV-associated lipodystrophy. Any other use, including body composition or fitness goals, is off-label.

Raising growth hormone is not free of risk: it can affect glucose regulation and fluid balance, which is one more reason this belongs under supervision with bloodwork. A side-by-side comparison is in sermorelin vs CJC-1295 vs tesamorelin.

None of these replaces the training itself, and none is a way to move a fitness age. The full range is in the compound catalogue and protocols.

Explore it with verified supply. Every Vivre Labs lot is independently tested and the certificate of analysis is published, so you know exactly what is in the vial.

Frequently asked questions

Is fitness age the same as biological age? No, and the difference is structural rather than a matter of accuracy. A fitness age is computed by estimating your cardiorespiratory fitness from heart rate at inferred workload plus the demographics you entered, then looking up the chronological age at which the average person of your sex in a reference population has that fitness value. It is a percentile expressed in years. A biological-age test, in the sense the term is normally used, is a statistical model built on a molecular layer - DNA methylation, or plasma proteins - fitted against chronological age, a mortality composite or a rate of decline. One reports how your aerobic capacity compares with a population; the other attempts to read a molecular state. They are not versions of the same measurement, and neither is a validated surrogate for healthspan or lifespan.

What does fitness age on a smartwatch actually mean? It means: the device's estimate of your cardiorespiratory fitness matches the average value for someone of that age and your sex, in whichever reference cohort the manufacturer used. That is the entire content of the claim. It does not describe your cells, your organs or your rate of ageing, and it depends on a reference population that is generally not disclosed - change the cohort and the number changes without you changing. It is also bounded: implementations floor and cap the output, so a very fit person can sit at the floor and stop receiving information.

How accurate is a wearable VO₂max estimate? The honest answer is that it is better than nothing at group level and weak at individual level, and this page makes no claim about any specific product. The estimate rests on an optically sensed heart rate, an external workload that is inferred rather than measured, and a maximal heart rate that is usually assumed from an age-based formula rather than measured - and an error in that assumption propagates straight through. Validation work generally reports acceptable mean agreement with laboratory cardiopulmonary exercise testing alongside wide individual limits of agreement, with regression toward the development population's mean: less-fit people over-estimated, highly fit people under-estimated. Individual errors are frequently large enough to move someone a whole fitness category. If you need a number you can act on, the laboratory test is the one that gives it to you.

Do readiness and recovery scores mean anything? Not as measurements. They are undisclosed weighted composites of resting heart rate, heart-rate variability, estimated sleep and recent training load; the weights differ between manufacturers and change with software updates, and they have not been validated against health outcomes. As behavioural prompts - a nudge to train lighter after a poor night - they can be useful, and there is no harm in that. As physiological readouts they should not be used, and they have nothing to say about biological age. Heart-rate variability specifically is worth tracking only against your own baseline with an unchanged measurement method; there is no external reference range for it.

Related reading

• Why biological age tests disagree - training targets, reference populations and assay noise, and why not to average two results
• What biological age tests measure - the families of test, what each is built from, and why no clock is a validated surrogate endpoint
• Functional markers of biological age - VO₂max, grip strength and gait speed, and the gap between predicting an outcome and changing it
• CGMs for people without diabetes - the other consumer measurement that is genuinely informative and easy to over-read
• Part of Testing & Biological Age

---

Educational content, not medical advice. No wearable device, brand, model or manufacturer is named, ranked, compared or recommended anywhere on this page, and nothing here constitutes an accuracy claim about any specific device; statements about estimation error describe the published literature on the sensing modality and estimation approach as a category. No doses or schedules appear; compounds and protocols named in the physician-discussion section are linked for information, not prescribed. Neither a wearable fitness estimate nor an epigenetic or organ clock is a diagnostic test, and no such score is currently accepted by regulators as a validated surrogate endpoint for healthspan or lifespan. Nothing on this page should be used to self-assess cardiovascular fitness or health status, to decide whether to begin or intensify exercise, or in place of clinical assessment. Maximal or near-maximal exercise testing carries risk and should be undertaken with appropriate medical assessment. Any result that concerns you, and any measurement discussed here, should be interpreted by a clinician who knows your history. Claims for the Malaysian market are subject to MAB requirements. Position stated as of 15 September 2026.

References

• Nes BM, Janszky I, Vatten LJ, Nilsen TIL, Aspenes ST, Wisløff U. Estimating VO2peak from a nonexercise prediction model: the HUNT Study, Norway. Medicine & Science in Sports & Exercise. 2011;43(11):2024–2030. (The normative-reference approach behind the "fitness age" concept: estimated peak oxygen uptake mapped onto population age norms so that a percentile can be communicated as an age. The origin of the construction described in this article.)
• Kodama S, Saito K, Tanaka S, et al. Cardiorespiratory fitness as a quantitative predictor of all-cause mortality and cardiovascular events in healthy men and women: a meta-analysis. JAMA. 2009;301(19):2024–2035. (The outcome validation of the underlying quantity - measured fitness, not a wrist estimate of it.)
• Mandsager K, Harb S, Cremer P, Phelan D, Nissen SE, Jaber W. Association of cardiorespiratory fitness with long-term mortality among adults undergoing exercise treadmill testing. JAMA Network Open. 2018;1(6):e183605. (Graded relationship persisting across the observed fitness range; measured, laboratory-derived fitness.)
• Ross R, Blair SN, Arena R, et al. Importance of assessing cardiorespiratory fitness in clinical practice: a case for fitness as a clinical vital sign. Circulation. 2016;134(24):e653–e699. (American Heart Association scientific statement; also the standard reference for the difference between measured and estimated fitness, and for the limitations of non-exercise and submaximal estimation.)
• Bent B, Goldstein BA, Kibbe WA, Dunn JP. Investigating sources of inaccuracy in wearable optical heart rate sensors. npj Digital Medicine. 2020;3:18. (Optical photoplethysmography error attributable to motion, activity type and physiological characteristics including skin tone, assessed across multiple consumer devices. Cited here as a property of the sensing modality; no individual device is identified or compared in this article.)
• On age-predicted maximal heart rate: the standard age-based prediction equations carry a wide individual standard error, such that an assumed maximum can differ substantially from a measured one. Any fitness estimate that depends on an assumed maximum inherits that error. (General exercise-physiology literature; see the Circulation statement above for discussion.)
• On validation of consumer wrist-derived VO₂max estimation generally: published comparisons against laboratory cardiopulmonary exercise testing typically report acceptable group-level mean agreement together with wide individual limits of agreement, and systematic regression toward the development population's mean - over-estimation at lower measured fitness, under-estimation at higher. Performance degrades where external workload cannot be measured, for example in strength training, indoor and non-ambulatory activity. Individual device validations are not cited here because this article makes no claim about any named product.
• On heart-rate variability: reported values depend on the metric computed, the recording window, time of day, posture, respiration rate and measurement method; within-person day-to-day variation is large, and values obtained by different methods are not interchangeable. There is no established external reference range for consumer-derived values. Useful as a personal baseline under unchanged method; not an ageing measurement.
• On readiness, recovery and composite scores: proprietary weighted combinations of resting heart rate, heart-rate variability, estimated sleep and recent training load. Weighting is undisclosed, differs between manufacturers and changes with software revision. No such composite has been validated against a health outcome, and none is an ageing measurement.
• On epigenetic clock construction and generation: first-generation clocks (Horvath, Hannum) fitted against chronological age; second-generation (PhenoAge, GrimAge) against mortality or clinical-biomarker composites; DunedinPACE against the rate of longitudinal physiological decline rather than a level. The "years" unit is a presentation choice in all cases. See what biological age tests measure for the full treatment.
• On epigenetic clock reliability: Nature Aging 2022 reported deviations of several years between replicate runs of the same sample, with principal-component versions bringing most replicates within about a year; a 2025 bioRxiv analysis reports clocks moving between ordinary days with no intervention, which sets a floor on detectable change. A 2025 Nature Communications comparison of 14 clocks against 174 disease outcomes in 18,859 people placed second-generation clocks well ahead of first-generation. Summarised as cited in the sibling articles linked above.
• On surrogate endpoints: no epigenetic clock, organ clock or composite biological-age measure is currently accepted by regulators as a validated surrogate endpoint for healthspan or lifespan, and no wearable-derived score has such status either.