Biological age, fitness age, and what Titan Age can tell you

Between their 20s and 30s, healthy adults lose 3 to 6% of their peak aerobic capacity per decade, measured in the same people over a median of 7.9 years (Fleg, 2005). By their 70s the loss exceeds 20% per decade. Set that against the best accuracy figure anyone has published for a consumer fitness age. It is plus or minus 3 to 5 years for Garmin's Fitness Age under good outdoor running conditions, a number from an enthusiast blog summarizing the vendor's own validation work, with no independent audit behind it. A single reading can carry as much error as a decade of real aging produces in a healthy 35-year-old.

An age estimate from a phone and a watch is a fitness and recovery signal converted into years. It says nothing about the state of your cells. This piece places wearable estimates among the four methods researchers use to estimate biological age, walks through the Norwegian HUNT fitness age model, puts numbers on how fast a fitness age can realistically move, and explains why devices disagree. It ends with what Titan Age is built from and how to read it.

01Four ways to estimate a biological age

Biological age is a family of estimates, each built from a different sample and trained against a different target. Knowing which category a number belongs to tells you what it can support.

Epigenetic clocks

Horvath (2013) built the first multi-tissue clock from 8,000 samples across 82 datasets and 51 healthy tissues and cell types. It reads methylation at 353 CpG sites, and in held-out test data its predictions correlated with chronological age at 0.96, with a median absolute error of 3.6 years. The clock was trained to predict chronological age. That makes its errors the interesting part. A person whose DNA methylation age runs ahead of their birth certificate has "age acceleration," and Horvath found every one of 20 cancer types showed it, by an average of 36 years.

Levine and colleagues (2018) changed the training target. They first built a clinical measure of phenotypic age from blood chemistry, then trained a methylation predictor on 513 CpGs to match it. The result, DNAm PhenoAge, predicted all-cause mortality, cancers, healthspan, physical function and Alzheimer's disease better than the first-generation clocks.

Belsky and colleagues (2022) changed the question again. The Dunedin Study has followed all 1,037 babies born in Dunedin, New Zealand in 1972 and 1973. The researchers tracked 19 biomarkers of cardiovascular, metabolic, renal, hepatic, immune, dental and pulmonary function at ages 26, 32, 38 and 45, and modeled how fast each person was declining. They then compressed that two-decade trajectory into a single blood test that reads 173 CpG sites. DunedinPACE is scaled so that 1.0 means one year of biological aging per calendar year. It asks how fast you are aging now, the same question consumer "pace of aging" readouts ask, Titan's included.

All three need a blood or saliva sample processed on a methylation array in a lab. No wearable can read DNA methylation, a point Dalia Health's comparison of consumer apps makes without qualification.

Blood biomarker clocks

The first step of Levine's work is itself a biological age method that needs no methylation data. Phenotypic age combines chronological age with nine standard lab values: albumin, creatinine, glucose, C-reactive protein, lymphocyte percentage, mean cell volume, red cell distribution width, alkaline phosphatase and white blood cell count. The weights come from a mortality model, so the output is the age at which your mortality risk would be average.

Cleveland Clinic describes the same family of methods more broadly, as equations that combine blood pressure, blood sugar, cholesterol, heart rate, height and weight into one number. The clinic notes that these estimates are widely used in aging research, that the reliability of consumer tests is unclear, and that none are part of routine medical care. A blood biomarker clock needs a venous draw and a lab, though a few of its inputs, such as heart rate and body size, are things a device can record.

Fitness age from cardiorespiratory fitness

Fitness age starts from VO2 max, the maximal rate at which your body takes in and uses oxygen. You find the age at which your VO2 max would be typical for your sex, and that age is your fitness age. Garmin describes its Fitness Age the same way. It "compares your current VO2 max fitness level to the normal values of different ages of people (of the same gender)."

Most consumer wearable age numbers belong to this category or borrow from it. The mortality evidence behind it is strong. In 122,007 adults followed for a median of 8.4 years after a treadmill test, those in the bottom quarter of fitness for their age and sex died at 5.04 times the adjusted rate of the elite group (Mandsager, 2018). Across 33 cohort studies and 102,980 participants, each 1-MET higher level of fitness, about 3.5 mL/kg/min, carried a 13% lower risk of all-cause death (Kodama, 2009). The American Heart Association's scientific statement (Ross, 2016) argued that fitness should be assessed in clinical practice as a vital sign, because it may predict mortality more strongly than smoking, hypertension, high cholesterol or type 2 diabetes.

A fitness age inherits that predictive weight and none of the epigenetic machinery. It tells you how your aerobic capacity compares with other people's, expressed in years.

Heart rate based estimates

The simplest methods skip a VO2 max model and work from heart rate alone. Uth and colleagues (2004) derived the heart rate ratio method from the Fick principle: VO2 max in mL/kg/min is about 15.3 times the ratio of maximal to resting heart rate. In 46 well-trained men, the formula estimated VO2 max with a standard error of 2.7 mL/kg/min when maximal heart rate was measured, and 4.7 mL/kg/min when it came from an age formula. The authors noted it would need separate validation in other groups.

The arithmetic shows why these estimates are fragile. With a maximal heart rate of 180 bpm, a resting heart rate of 60 gives 45.9 mL/kg/min and a resting heart rate of 55 gives 50.1. Five beats of resting heart rate, the difference between a good night and a night after alcohol, moves the estimate by more than 4 mL/kg/min. The meta-analysis by Molina-Garcia and colleagues (2022) found the same weakness in consumer devices. Wearables that estimated VO2 max from resting data overestimated it by 2.17 mL/kg/min on average, with limits of agreement from 13.07 below to 17.41 above. Devices that used exercise data showed almost no bias, and their limits of agreement still ran from about 9.9 below to 9.7 above.

Heart rate recovery is the other heart rate signal with mortality data. Cole and colleagues (1999) followed 2,428 adults for six years after an exercise test. A drop of 12 bpm or less in the first minute after peak exercise carried twice the adjusted risk of death, independent of workload, and the authors read the slow recovery as a sign of reduced vagal activity. Heart rate recovery says something real about autonomic function. It is a risk marker, and converting it to years adds a modeling step with no validation study behind it.

02How the HUNT fitness age model works

Most people have never had a VO2 max test. The Norwegian HUNT study solved that problem with a regression model, and NTNU's public fitness age calculator is built on it. The method runs in three steps.

Step one, predict VO2 peak without a test

Nes and colleagues (2011) measured VO2 peak directly on a treadmill in 4,637 healthy adults from the HUNT population. They built separate models for 2,067 men with a mean age of 48.8 and 2,193 women with a mean age of 47.9, then cross-validated each on a split sample. Four predictors carried the model, in this order of strength for both sexes: age, waist circumference, leisure-time physical activity, and resting heart rate. The activity input is a self-reported index of how often, how long and how hard you exercise.

The models explained 61% of the variance in measured VO2 peak for men and 56% for women. The standard error of the estimate was 5.70 mL/kg/min for men and 5.14 for women. The authors called the result "fairly accurate" and suitable for "a rough assessment of cardiorespiratory fitness in an outpatient setting," which is a fair description. About 40% of the variation between people lies outside what age, waist, activity and pulse can explain.

NTNU's calculator asks for sex, age, height, weight, waist circumference or BMI, how often and how hard you exercise, and resting heart rate, and cites the 2011 paper as its source.

Step two, convert fitness into an age

The predicted VO2 peak is then compared with population values by age and sex. If your predicted fitness matches the median for people 15 years younger, your fitness age is 15 years below your chronological age. This lookup step is where every product makes its own choices. HUNT's reference values come from Norwegian adults. The FRIEND registry (Kaminsky, 2015) provides the US reference, from 7,783 adults aged 20 to 79 tested on a treadmill. The VO2 max guide has the full FRIEND tables by age decade and sex, with two worked examples of this median matching.

The lookup has a shape worth knowing. In the FRIEND data, the median VO2 max for men falls by roughly 5 mL/kg/min per decade between the 20s and the 60s, about 0.5 mL/kg/min per year of age. For women the drop is steepest between the 20s and 30s and runs closer to 3.4 mL/kg/min per decade from the 30s through the 60s. Every mL/kg/min of error in step one therefore becomes about two years of fitness age for a man in midlife, and about three for a woman.

Step three, check the prediction against mortality

A fitness estimate that nobody tested against outcomes would be a curiosity. Nes and colleagues (2014) applied the 2011 model to 37,112 healthy adults from the first HUNT survey, in 1984 to 1986, and followed them for a mean of 24 years. Among participants under 60 at baseline, each 1-MET higher estimated fitness was associated with 21% lower cardiovascular mortality in both sexes, and 15% lower all-cause mortality in men and 8% in women.

The more instructive result is a comparison. The fitness estimate discriminated mortality risk with an area under the curve of 0.70 to 0.77. Each of its inputs alone scored 0.55 to 0.63, and a plain sum of the inputs' standardized scores scored 0.61 to 0.65. The same four numbers predicted death better when they were weighted by a model trained on measured oxygen uptake. The weighting carries information. Any age estimate built from wearable inputs makes its own weighting choices, and unless those weights were trained against an outcome, the estimate has no comparable evidence behind it.

Where maximal heart rate fits

Nes and colleagues (2013) tested 3,320 healthy adults to a verified maximal effort and found maximal heart rate followed 211 minus 0.64 times age, with no meaningful interaction with sex, activity, fitness or BMI. Older formulas, including 220 minus age, underestimated measured maximum in people over 30. At age 50 the HUNT formula gives 179 bpm and 220 minus age gives 170.

The standard error was 10.8 bpm. Any method that needs maximal heart rate carries that error unless the maximum is observed. Titan's heart rate zones use the higher of your highest heart rate in Apple Health over the last 30 days and 220 minus your age, so a hard effort above the formula replaces it with your own number.

03What moves a fitness age and by how much

The HUNT model points to four levers. Each one moves a fitness age at a different speed, and knowing the speeds is the best defense against over-reading a weekly change.

VO2 max. Measured fitness is the direct input, and it responds to training faster than it declines with age. In a meta-analysis of 41 controlled trials in 2,102 sedentary adults aged 60 and older, endurance training raised VO2 max by 3.78 mL/kg/min, or 16.3%, over control groups (Huang, 2005). Programs longer than 20 weeks at roughly 60 to 70% of VO2 max produced the largest gains. In the FRIEND medians, the gap between men in their 60s and men in their 70s is 3.8 mL/kg/min. On average, about five months of training returned the fitness that separates one decade of age from the next.

Resting heart rate. Resting heart rate falls with aerobic training and rises with poor sleep, alcohol, heat, illness and accumulated fatigue. It is also the input most likely to move for reasons unrelated to fitness. A few beats of change over a week is normal day-to-day variation, and the Uth arithmetic above shows how far that can push a heart rate based estimate.

Waist and body composition. Waist circumference ranked second among the HUNT predictors. Body mass also enters any VO2 max expressed per kilogram, so losing fat raises relative VO2 max even when the heart and muscles deliver the same oxygen per minute. A fitness age can improve after weight loss with no change in aerobic capacity measured in liters per minute.

Activity. The self-reported activity index in HUNT stands in for fitness the model cannot measure. On a wearable, recorded activity replaces the questionnaire. Activity moves week to week. The fitness it builds takes months.

Real change against week-to-week noise

The table puts the pace of physiology next to the size of measurement error, using the FRIEND medians for men to convert mL/kg/min into years.

Two things follow. A man of 35 at the FRIEND median of 42.4 mL/kg/min loses somewhere between 1.3 and 2.5 mL/kg/min over the next ten years by Fleg's longitudinal rate. That is a few thousandths of a unit per week. When a fitness age moves five years in one week, which for a man in midlife means about 2.5 mL/kg/min, aerobic capacity did not change that much. The estimate did. A new outdoor run in cooler weather, a device update, a changed body weight entry, or a resting heart rate pushed up by a bad night are the usual causes.

Training, by contrast, produces real change on a scale of months. A fitness age trend that falls steadily across a 16 to 20 week block of consistent aerobic work, while pace at a fixed heart rate improves, is describing something that happened.

04How accurate a wearable age can be

The Garmin figure is the only stated accuracy number for a consumer fitness age. The5krunner, summarizing Firstbeat's validation studies, reports a mean absolute VO2 max error near 3.5 mL/kg/min under controlled outdoor conditions with a chest strap, and translates it into plus or minus 3 to 5 years of Fitness Age. The source is a single blogger's reading of vendor claims. No peer-reviewed study has audited it.

Checked against the FRIEND medians, the translation looks generous. For a man in midlife, 3.5 mL/kg/min spans about seven years of median-matched age, and for a woman in her 40s or 50s it spans about a decade. Garmin may use reference tables with a different slope, and the vendor does not publish them. Using the published US norms, the 3 to 5 year figure sits at the optimistic end of what the stated VO2 max error allows.

The independent data are wider. Lambe and colleagues (2025) compared Apple Watch estimates with a maximal treadmill test in 30 adults. The watch read low by 6.07 mL/kg/min on average, with a mean absolute error of 6.92 mL/kg/min and limits of agreement from 6.11 above the measured value to 18.26 below. The HUNT regression, built with a tape measure and a questionnaire, has a standard error of 5.1 to 5.7 mL/kg/min. Using the conversion above, an error of that size is worth ten or more years of fitness age.

That is the ceiling for any single wearable-derived age reading. Its precision is measured in decades. Part of that error is a systematic offset, since Apple Watch read low on average in Lambe's study, and an offset cancels out when a device is compared with itself over months. Dalia Health's review of consumer apps says a wearable estimate "can easily be off by a few years, and the number will bounce around with your sleep, stress, training, and even how you measured." Use the trend. Treat the absolute number as a rough placement.

05Why every wearable gives you a different number

Two devices on the same wrist in the same week can disagree by years and both be working as designed. They measure different things and convert them with different reference tables.

Apple stops at the fitness estimate. Apple Watch records VO2 max on outdoor walks, runs and hikes across a supported range of 14 to 65 mL/kg/min and classifies it against your age and sex for users 20 and older. It never converts the number to years.

Garmin takes the next step. Its Fitness Age is a fitness age in the HUNT sense, with the inputs moved from a questionnaire to the wrist.

Oura measures something else. Cardiovascular Age estimates pulse wave velocity, the speed at which the pulse travels through the arteries, from features of the ring's optical signal. Stiffer arteries carry the pulse faster. Oura's development study had 600 participants, run with the Kuopio Research Institute of Exercise Medicine and UCLA, and Oura states that its products are not medical devices. Arterial stiffness and aerobic capacity are separate properties. A person with a high VO2 max and stiff arteries could get a young Garmin age and an old Oura age in the same week, and both could be correct.

WHOOP, as Dalia Health describes it, estimates WHOOP Age from about six months of data across nine metrics: VO2 max, resting heart rate, lean body mass, sleep duration and consistency, daily steps, time in heart rate zones, and strength activity. A Pace of Aging score from -1x to 3x shows whether recent habits are speeding the number up or slowing it down. The weighting is not published.

The disagreements follow from the design choices. A fitness age moves with VO2 max. A vascular age moves with arterial stiffness. A composite moves with sleep, steps and training volume as well as fitness, and each vendor weights those differently and smooths them over a different window. Compare each number with its own history.

06What fitness age and epigenetic age share

The two ends of the taxonomy do point the same way, weakly. Kawamura and colleagues (2026) measured VO2 peak and DunedinPACE in 144 Japanese men aged 65 to 72. Higher VO2 peak was associated with slower epigenetic aging, with a correlation of -0.16 after adjusting for age, smoking and drinking. A VO2 peak of 26.2 mL/kg/min best separated men with slower DunedinPACE from the rest.

The size of that correlation is the useful part. A correlation of -0.16 means fitness accounts for less than 3% of the variation in epigenetic pace of aging in that sample. The design was cross-sectional, so it cannot show that raising VO2 max slows methylation aging, and the authors say so. Fitness and epigenetic age are related measures of different biology. A good fitness age is strong evidence about mortality risk, and weak evidence about what a methylation array would report.

07What Titan Age is built from

Titan Age sits in the composite category, next to WHOOP Age. It starts from your actual age at the end of the last completed Monday to Sunday week, read from the date of birth in Apple Health, and adds or subtracts years for each of 11 signals. Sex is not an input. In the help article's own words, Titan Age "is a model estimate and does not measure your body's age."

The 11 signals fall into three groups. Recovery carries 46% of the model: resting heart rate 18%, sleep duration 12%, heart rate recovery 10%, and sleep consistency 7%. Activity carries 35%: weekly steps 9%, Zone 2 minutes 8%, Zone 4 and above minutes 8%, strength training minutes 5%, and stand hours 5%. Fitness carries 19%: VO2 max 16% and lean mass index 3%. The personalized metrics reference lists the same signals in one paragraph.

Compared with the HUNT model, Titan Age puts far less weight on fitness and far more on recovery and weekly behavior. VO2 max, the quantity a classic fitness age exists to predict, is 16% of the model. Resting heart rate appears in both. Heart rate recovery starts from Apple Watch's one-minute value, the interval Cole studied, which Titan multiplies by 1.65 to estimate a two-minute drop.

HRV is not one of the 11 signals. Titan uses HRV for the Recovery score, which compares today's HRV and resting heart rate with the median of your baseline window of 7, 30 or 60 days, in units of your own median absolute deviation. Vitals shows HRV against a personal 20th to 80th percentile range, which is not a clinical reference range.

Each signal converts to years through a reference value, a step size, the signal's weight and a coverage factor. For resting heart rate the reference is 80 bpm and the step is 5 bpm, so a week averaging 60 bpm lowers Titan Age by about 1.7 years at full coverage. Missing signals lower coverage, and with little history they add zero years, so missing data pulls Titan Age toward your actual age.

Titan Age moves slowly by design. Each week the displayed value moves between 8% and 62% of the way from last week's Titan Age toward the new result, faster with good coverage and a history of larger swings, slower with sparse data. Given how slowly aerobic fitness changes, that damping is the right choice. A five-year jump in one week is a data problem, and the damping keeps one unusual week from moving the headline by its full amount.

Pace of aging is a separate readout. With fewer than 12 weeks of history, it is the gap between Titan Age and your actual age as a share of your actual age, so a Titan Age of 35 at age 41 reads -14.6%. From 12 weeks on, Titan takes up to 52 weeks of Titan Age, lightly averages the week-to-week changes, converts them to years of change per year, and expresses that as a percent of your current Titan Age. A Titan Age falling 0.02 years each week at 35 is about -1 year per year, or -3.0%. Negative is the favorable direction in both cases. It asks the same question as DunedinPACE, how fast you are aging now, from very different data. DunedinPACE reads methylation in blood. Titan's pace reads trends in your own recorded behavior and physiology.

Titan Age has not been validated against a lab VO2 max test, an epigenetic clock, a blood biomarker clock, or mortality, and no accuracy figure exists for it. Titan has no epigenetic testing and no blood panel, and the Garmin accuracy figure above applies to Garmin only. What Titan Age does offer is a consistent weekly summary on one scale, with the contribution of each signal shown in years. Several of its inputs, such as VO2 max and heart rate recovery, carry outcome evidence of their own. Read it as a record of whether your recorded recovery, activity and fitness are moving in the right direction, and read the pace as the speed of that movement.

08The Titan Age screen

The Titan Age card on Today names the week it covers in its top right corner. Tap it to open the Titan Age screen. Select any week on the Age history chart to see Titan Age and Pace for that week. Tap any signal row to see its Age contribution in years, and open Calculation details to see each signal's value, target, weight, and days of coverage.

Start with coverage. A signal short of its minimum days, such as 4 nights of sleep or 4 days of resting heart rate, counts as missing, and missing signals shrink every other signal's effect. Then look at the top three weekly movers, the signals whose contribution changed most from the previous week. A negative mover pulled Titan Age down. If a mover looks implausible, check the value in Calculation details before changing anything about your training. Current-week workouts, steps and sleep wait until Sunday ends. If the card is blank or stuck, Titan Age missing walks through the date of birth and weekly update checks.

Judge Titan Age by its trend across a training block, and judge pace once 12 weeks of history exist. Neither number is a measurement of your cells.

09References

1. Apple (n.d.). Track your cardio fitness levels. Apple Support. https://support.apple.com/en-us/108790
2. Belsky DW et al. (2022). DunedinPACE, a DNA methylation biomarker of the pace of aging. eLife. https://doi.org/10.7554/eLife.73420
3. Cleveland Clinic (n.d.). What is biological age? Cleveland Clinic Health Essentials. https://health.clevelandclinic.org/biological-age
4. Cole CR et al. (1999). Heart-rate recovery immediately after exercise as a predictor of mortality. New England Journal of Medicine. https://doi.org/10.1056/NEJM199910283411804
5. Dalia Health (n.d.). How apps calculate biological age. Dalia Health blog. https://daliahealth.com/blog/how-apps-calculate-biological-age
6. Fleg JL et al. (2005). Accelerated longitudinal decline of aerobic capacity in healthy older adults. Circulation. https://doi.org/10.1161/CIRCULATIONAHA.105.545459
7. Garmin (n.d.). What is your Garmin Fitness Age? How can you lower it? Garmin blog. https://www.garmin.com/en-US/blog/fitness/what-is-your-garmin-fitness-age-how-can-you-lower-it/
8. Horvath S (2013). DNA methylation age of human tissues and cell types. Genome Biology. https://doi.org/10.1186/gb-2013-14-10-r115
9. Huang G et al. (2005). Controlled endurance exercise training and VO2max changes in older adults: a meta-analysis. Preventive Cardiology. https://doi.org/10.1111/j.0197-3118.2005.04324.x
10. Kaminsky LA et al. (2015). Reference standards for cardiorespiratory fitness measured with cardiopulmonary exercise testing: data from the Fitness Registry and the Importance of Exercise National Database. Mayo Clinic Proceedings. https://doi.org/10.1016/j.mayocp.2015.07.026
11. Kawamura T et al. (2026). Revisiting the relationship between cardiorespiratory fitness and biological aging: insights from DunedinPACE analysis. GeroScience. https://doi.org/10.1007/s11357-025-01930-0
12. Kodama S et al. (2009). Cardiorespiratory fitness as a quantitative predictor of all-cause mortality and cardiovascular events in healthy men and women: a meta-analysis. JAMA. https://doi.org/10.1001/jama.2009.681
13. Lambe R et al. (2025). Investigating the accuracy of Apple Watch VO2 max measurements: a validation study. PLOS One. https://doi.org/10.1371/journal.pone.0323741
14. Levine ME et al. (2018). An epigenetic biomarker of aging for lifespan and healthspan. Aging. https://doi.org/10.18632/aging.101414
15. Mandsager K et al. (2018). Association of cardiorespiratory fitness with long-term mortality among adults undergoing exercise treadmill testing. JAMA Network Open. https://doi.org/10.1001/jamanetworkopen.2018.3605
16. Molina-Garcia P et al. (2022). Validity of estimating the maximal oxygen consumption by consumer wearables: a systematic review with meta-analysis and expert statement of the INTERLIVE network. Sports Medicine. https://doi.org/10.1007/s40279-021-01639-y
17. Nes BM et al. (2011). Estimating VO2peak from a nonexercise prediction model: the HUNT Study, Norway. Medicine and Science in Sports and Exercise. https://doi.org/10.1249/MSS.0b013e31821d3f6f
18. Nes BM et al. (2013). Age-predicted maximal heart rate in healthy subjects: the HUNT Fitness Study. Scandinavian Journal of Medicine and Science in Sports. https://doi.org/10.1111/j.1600-0838.2012.01445.x
19. Nes BM et al. (2014). A simple nonexercise model of cardiorespiratory fitness predicts long-term mortality. Medicine and Science in Sports and Exercise. https://doi.org/10.1249/MSS.0000000000000219
20. NTNU Cardiac Exercise Research Group (n.d.). Fitness calculator. NTNU. https://www.ntnu.edu/cerg/vo2max
21. Oura (n.d.). All about Cardiovascular Age. Oura blog. https://ouraring.com/blog/cardiovascular-age/
22. Ross R et al. (2016). Importance of assessing cardiorespiratory fitness in clinical practice: a case for fitness as a clinical vital sign. A scientific statement from the American Heart Association. Circulation. https://doi.org/10.1161/CIR.0000000000000461
23. the5krunner (n.d.). Garmin Fitness Age: the Firstbeat algorithm and its accuracy. the5krunner. https://the5krunner.com/garmin-features/physiology/fitness-age/
24. Uth N et al. (2004). Estimation of VO2max from the ratio between HRmax and HRrest, the Heart Rate Ratio Method. European Journal of Applied Physiology. https://doi.org/10.1007/s00421-003-0988-y