The honest version of "biological age"

Biological age is a statistical estimate of how old your body acts compared to people of your chronological age. It is built from biomarkers and behavioural inputs that are known, from cohort studies, to track with disease risk and mortality. The output looks like a single tidy number - "biological age: 38" against a chronological age of 45 - but the math underneath is a regression model, and the number is an estimate with a confidence interval the model rarely shows you.

This matters because the marketing tends to flatten the uncertainty. The honest framing is: biological age is a useful direction-of-travel signal built from several real, peer-reviewed biomarkers. The composite number you see on a screen is less precise than the underlying biomarkers it is built from. If three independent methods all agree you are aging slowly, you probably are. If they disagree - and they often do - the underlying biomarkers tell the truer story than any single composite.

The rest of this article walks through the four most common inputs (VO₂ max, resting heart rate, HRV, and the blood-based epigenetic clocks), what each one is actually measuring, and how well each one predicts what we care about: how long you live, and how long you live well.

VO₂ max: the strongest single number we have

VO₂ max is the maximum amount of oxygen your body can consume during peak exertion, measured in millilitres per kilogram per minute. It is a direct readout of cardiorespiratory fitness - how well your lungs, heart, vasculature, and mitochondria work together under load. Of all the non-genetic inputs that go into bio-age estimates, VO₂ max has the strongest mortality signal in the published cohort data.

The reference paper most longevity writing cites is Mandsager et al., JAMA Network Open, 2018 [1], a 122,007-patient observational study from the Cleveland Clinic. Moving from the lowest fitness quintile to even the next-lowest was associated with roughly a 50% reduction in all-cause mortality, sustained over a decade of follow-up. Moving from "low" to "elite" cut mortality risk by a factor of about five. The effect size was larger than the effect size of smoking, hypertension, or Type 2 diabetes - and it was dose-dependent (more fitness, more benefit, with no observed upper plateau).

How accurate is the measurement? On a wrist-worn wearable, VO₂ max estimates are derived from heart-rate response during sub-maximal activity, calibrated against population data. Validation studies (e.g. Klepin et al., Journal of Sports Sciences, 2022 [2]) typically find a standard error of around 3.5–5 ml/kg/min versus gas-exchange ground truth - which sounds large until you realise the typical year-on-year change a healthy adult is trying to detect is around 2 ml/kg/min. That is why month-to-month wobble on Apple Watch or Garmin is mostly noise; the six-month trend is the signal.

The bottom line: VO₂ max is the single biomarker with the strongest direct link to lifespan, but the wearable-derived number is noisy at short time horizons. Trust the six-month trajectory, not yesterday's reading.

Resting heart rate: cheap, surprisingly powerful

Resting heart rate (RHR) is the number of beats per minute when you are fully at rest, ideally measured in the same conditions each day. It is the simplest cardiovascular biomarker we have, and it is more predictive than its simplicity suggests.

The Copenhagen Male Study (Jensen et al., Heart, 2013 [3]) followed roughly 3,000 healthy middle-aged men for 16 years. Each 10-bpm increase in RHR was associated with a 16% increase in all-cause mortality, independent of fitness, smoking, and blood pressure. The Norwegian HUNT-1 cohort (Nauman et al., JAMA, 2011 [4]) found similar magnitudes in both sexes across a 12-year follow-up: an RHR above 80 bpm carried roughly twice the mortality risk of an RHR below 65 bpm, even after adjusting for known confounders.

The mechanism is plausible and physiological: a lower RHR generally reflects higher vagal tone, larger stroke volume, and a more efficient cardiovascular system. A high RHR can reflect deconditioning, chronic inflammation, hidden infection, sleep debt, dehydration, or autonomic dysfunction - all of which independently predict worse outcomes.

Wearables measure RHR with much higher accuracy than VO₂ max. Photoplethysmography-based estimates (the green LEDs on the back of every wrist device) typically come within 1–3 bpm of an ECG ground truth at rest. The catch is methodological: most wearables report a 24-hour minimum or an average of the lowest sleep window, not a true clinical RHR taken in the morning at rest. The number is internally consistent, which means trends are reliable, but it should not be compared directly to a number taken in a doctor's office.

For bio-age purposes, the consensus from the cohort literature is that an RHR in the high 50s to low 60s is roughly the "young cardiovascular system" band for adults. Above 75, the mortality curves bend upward; below 50 in non-athletes can sometimes flag a different problem (e.g. a conduction abnormality) and warrants a check.

HRV: useful for recovery, weaker for bio age

Heart-rate variability - the millisecond-level variation in the interval between consecutive beats - is the noisiest of the four biomarkers we are walking through, and the one most subject to over-interpretation. Marketing around HRV often implies that a higher number is unambiguously better and that day-to-day change reflects underlying biological age. The peer-reviewed literature is more cautious on both counts.

What HRV is genuinely good at: tracking acute autonomic load. A meta-analysis by Sammito and Böckelmann (Heart Rhythm, 2016 [5]) confirmed that resting HRV (specifically the RMSSD metric) drops measurably under sleep deprivation, illness, training overload, and high alcohol intake. As a daily recovery signal, the 7-day rolling average is solid.

What HRV is weaker at: predicting mortality directly when other inputs are controlled. The ARIC cohort (Dekker et al., Circulation, 2000 [6]) found a real but modest association between low HRV and all-cause mortality, but the effect substantially attenuated once VO₂ max and RHR were added to the model. In other words, HRV mostly tells you the same story as fitness and resting heart rate, plus some short-term recovery noise on top.

Two practical implications for bio-age estimates: first, treat any bio-age model that weights single-day HRV heavily with skepticism - that is a model picking up day-to-day noise as if it were signal. Second, the 30-day rolling average of HRV, paired with RHR and VO₂ max trend, is a credible "cardiovascular age" panel. If you want the longer version on what wearable HRV is actually measuring and why the day-to-day reading is so noisy, we wrote about it here.

Epigenetic clocks: the closest to a "real" bio age

The most rigorously validated bio-age methods are not wearable-derived. They are blood-based assays that read the methylation pattern of your DNA at hundreds of specific sites and feed those values into a regression model trained against age and (in newer versions) mortality.

The current state of the art:

  • Horvath clock (2013) [7]: the first pan-tissue clock. Predicts chronological age within ~3.6 years across most tissue types. Useful as a baseline but not optimised for predicting health outcomes.
  • PhenoAge (Levine et al., 2018) [8]: built specifically to predict mortality and morbidity rather than chronological age. Each year of accelerated PhenoAge is associated with a roughly 10% increase in all-cause mortality risk over a decade of follow-up.
  • GrimAge (Lu et al., 2019) [9]: trained on plasma proteins and smoking history in addition to methylation. Currently the strongest single-measurement predictor of time-to-death in the major validation cohorts.
  • DunedinPACE (Belsky et al., 2022) [10]: measures the pace of biological aging (how much biological time is passing per chronological year), not the cumulative deficit. Particularly responsive to intervention studies because it can detect a slowdown in months rather than years.

How accurate are they? GrimAge and DunedinPACE both predict all-cause mortality with hazard ratios in the 1.10–1.15 range per accelerated year, across multiple independent cohorts. This is a real and useful signal. The two limitations to keep in mind: first, the test-retest reliability of any single methylation clock is in the 0.6–0.9 range - meaning if you measured yourself twice in a week, the two numbers might differ by 1–2 years just from measurement variance. Second, lifestyle interventions have been shown to move PhenoAge and DunedinPACE in randomised trials (e.g. Fitzgerald et al., Aging, 2021 [11]), but the effect sizes are modest - typically 1–3 years of bio-age reduction over 8 weeks of a structured diet, sleep, and exercise protocol.

For most people, a blood-based clock is worth doing once as a baseline and again 12 months later if you are running a serious intervention. More frequent re-testing tracks measurement noise more than real change.

How a wearable bio age is actually computed

The bio-age number on your wearable (or on Thier, for that matter) is a weighted composite of the inputs above plus some others - typically VO₂ max, RHR, HRV, sleep duration, weight, body composition if available, and sometimes activity volume and blood-pressure if you log it. Different apps weight the inputs differently; there is no consensus formula. The composite is then mapped onto reference data from large cohort studies so the output reads as a number of years.

What that means in practice: the composite number is built to be intuitive (everyone understands "your body acts like a 38-year-old"), but the precision is bounded by the noisiest underlying input. If your VO₂ max estimate has a ±3-year band and your HRV trend has a ±1-year band, the composite cannot be more precise than the inputs - typically ±2–4 years.

This is why Thier reports both the composite number and the per-domain breakdown. The breakdown is where the actionable information lives. A composite of "bio age 42" against a chronological age of 45 is interesting; learning that your cardiovascular sub-score is 38 but your sleep sub-score is 49 tells you exactly where the next month of effort should go.

What "accurate" actually means here

There are three different things "accuracy" can mean when people talk about biological age, and they collapse together in most marketing:

  1. Reproducibility - if I measure the same person twice in a week, how close are the two numbers? For most wearable bio ages this is within ±2–3 years. For blood-based clocks it depends on the clock (Horvath ~3.6 years; GrimAge ~1.5 years).
  2. Construct validity - does the number actually correlate with the underlying biology it claims to measure? For VO₂ max and RHR, yes, strongly. For methylation clocks, also yes, in large validation cohorts. For composite bio-age numbers, indirectly - the inputs are valid, the weights are best-guess.
  3. Predictive validity - does the number predict outcomes that matter (mortality, disease, healthspan)? The strongest evidence is for VO₂ max (mortality hazard ratios above 4x between fitness extremes), then RHR, then GrimAge and DunedinPACE. The composite numbers inherit predictive validity from their inputs.

When a marketing page says a bio-age product is "95% accurate", they are almost always quoting reproducibility, which is the least interesting of the three. The number that should drive what you actually do is predictive validity - and that is where the underlying biomarkers, not the composite, are the right thing to track.

Practical advice: how to use the number well

If you have any kind of bio-age readout - wearable-derived, app-derived, or from a methylation test - here is how to extract real value from it without overreading the precision:

  • Watch the trend, not the day. A six-month moving average of your bio age is far more informative than this morning's reading. Day-to-day change is almost entirely noise.
  • Trust the inputs over the composite. VO₂ max, RHR, and HRV trend each have stronger published validity than the composite that wraps them. If your VO₂ max is climbing steadily, the bio-age number is going the right way even if the composite hides it for a month or two.
  • Don't compare across methods. A wearable bio age of 38 is not the same scale as a DunedinPACE bio age of 38. Pick one method, watch the trend on that method.
  • Recalculate monthly at most. For lifestyle decisions, a monthly check is plenty. For blood-based clocks, 6–12 months between tests.
  • Use the per-domain breakdown to decide where to spend effort. If sleep is dragging your composite, that is where the next month of work goes - not on another supplement that promises to "lower biological age".

If you want a deeper dive on the specific levers that move these biomarkers in the right direction, the companion piece is our evidence-ranked playbook on lowering biological age in your forties. The TL;DR is that the same inputs do most of the work: aerobic base, strength, sleep, blood pressure, lipid management, and the boring stuff like social connection.

The takeaway

Biological age is a real, peer-reviewed concept built on biomarkers with genuine predictive validity. The composite number on a wearable or app is less precise than the marketing suggests, but the underlying inputs - VO₂ max, RHR, HRV trend, and (if you go further) a methylation clock - are some of the most powerful predictors of long-term health we have.

Use the number as a direction-of-travel signal, not a diagnostic. Trust the inputs over the composite. Trust the trend over the day. Recalculate sparingly. And spend your time on the things that move the inputs - which, as it turns out, are the same things that move every other longevity marker.

If you want the underlying biomarkers calculated, tracked, and explained from your existing wearable + HealthKit data, that is what Thier does - have a look at the app.

Frequently asked questions

Is biological age a real thing or marketing?

It is a real, peer-reviewed concept - but it is a composite estimate rather than a single measured value. Different methods (epigenetic clocks, biomarker panels, wearable-derived models) all correlate with mortality and disease risk, but they disagree with each other by several years for the same person. Treat any single number as a direction-of-travel signal, not a diagnostic.

Which is more accurate: a wearable bio age or a blood-based one?

Blood-based methods (especially DunedinPACE and PhenoAge) have stronger predictive validity for mortality in published cohorts. Wearable-derived bio ages are more useful for tracking week-to-week change in a single person because the inputs (VO₂ max, RHR, HRV) are responsive to training and sleep. The two approaches answer different questions.

How much can VO₂ max really move my bio age?

Moving from the lowest fitness quintile to even the second-lowest is associated with a roughly 50% reduction in all-cause mortality (Mandsager et al., JAMA Network Open, 2018). In bio-age model terms, the average untrained 45-year-old who follows a structured aerobic program for 12 months typically sees their VO₂-derived sub-score drop by 3–7 years.

Is HRV a good way to track biological age?

HRV correlates with autonomic function and recovery, but as a single-time-point bio-age signal it is too noisy to be useful. The 30-day rolling average, viewed alongside RHR and VO₂ max, is a far better readout of cardiovascular age than any single morning's number.

How often should I recalculate my biological age?

For lifestyle decisions, once a month is plenty. The underlying biomarkers (especially HRV and RHR) move week to week with sleep, hydration, and stress, so daily recalculation tracks noise more than signal. Blood-based clocks should be repeated no more often than every 6–12 months.

References

  1. Mandsager K, Harb S, Cremer P, et al. Association of Cardiorespiratory Fitness With Long-term Mortality Among Adults Undergoing Exercise Treadmill Testing. JAMA Network Open. 2018;1(6):e183605. PubMed
  2. Klepin K, Wing D, Higgins M, et al. Validity of Cardiorespiratory Fitness Measured with Fitbit Compared to VO₂ max. Journal of Sports Sciences. 2022. PubMed
  3. Jensen MT, Suadicani P, Hein HO, Gyntelberg F. Elevated resting heart rate, physical fitness and all-cause mortality: a 16-year follow-up in the Copenhagen Male Study. Heart. 2013;99(12):882-887. PubMed
  4. Nauman J, Janszky I, Vatten LJ, Wisløff U. Temporal changes in resting heart rate and deaths from ischemic heart disease. JAMA. 2011;306(23):2579-2587. PubMed
  5. Sammito S, Böckelmann I. Reference values for time- and frequency-domain heart rate variability measures. Heart Rhythm. 2016. PubMed
  6. Dekker JM, Crow RS, Folsom AR, et al. Low heart rate variability in a 2-minute rhythm strip predicts risk of coronary heart disease and mortality. Circulation. 2000;102(11):1239-1244. PubMed
  7. Horvath S. DNA methylation age of human tissues and cell types. Genome Biology. 2013;14(10):R115. PubMed
  8. Levine ME, Lu AT, Quach A, et al. An epigenetic biomarker of aging for lifespan and healthspan. Aging. 2018;10(4):573-591. PubMed
  9. Lu AT, Quach A, Wilson JG, et al. DNA methylation GrimAge strongly predicts lifespan and healthspan. Aging. 2019;11(2):303-327. PubMed
  10. Belsky DW, Caspi A, Corcoran DL, et al. DunedinPACE, a DNA methylation biomarker of the pace of aging. eLife. 2022;11:e73420. PubMed
  11. Fitzgerald KN, Hodges R, Hanes D, et al. Potential reversal of epigenetic age using a diet and lifestyle intervention: a pilot randomized clinical trial. Aging. 2021;13(7):9419-9432. PubMed