VO2max After 50: What Training Can Still Change

The number arrives without ceremony, tucked between your resting heart rate and last night's sleep score, and it has been drifting downward — not dramatically, not the way a bad lab value drifts, but steadily enough that you have started noticing it, and then noticing that you are noticing it, which is its own small indignity; and because the watch offers no explanation, only the number and a little downward arrow and a cheerful suggestion that you move more, you are left to supply the story yourself, which is more or less where most people past fifty first meet the problem of aerobic capacity: as an unasked-for verdict, delivered by a wrist, about how much of what you have lost is age and how much is just the last few winters of driving to places you used to walk.

Then there is the corroborating evidence, which is worse because it is not a number: the hill on the way home, the one that used to be scenery and is now an event. None of it feels like disease. It feels like weather. The honest question underneath the watch number is not how to get back what you had at forty, which may not be available, but the more answerable one — what part of this is age, what part is disuse, and what can training still change?

Here is the most defensible answer the evidence permits. Aerobic capacity declines with age in everyone, the decline accelerates with each decade rather than running at a constant rate, and it does so in people who exercise as well as in people who do not. Low fitness is among the most consistently observed markers of mortality risk in large cohorts — but those are cohorts, not experiments, and being fit is partly a consequence of being well. Training in previously sedentary older adults reliably raises measured aerobic capacity by a few millilitres per kilogram per minute: modest, replicated, and highly variable between individuals. And a five-year randomised trial of supervised exercise in 1,567 older adults did not demonstrate a mortality benefit over ordinary activity guidelines. Training moves the number, the number tracks risk, and the chain between them has not been closed by a trial.

The decline is real, and it does not proceed in a straight line

Almost everything you have absorbed about a steady, fixed annual percentage loss after thirty comes from cross-sectional data — measuring people of different ages once and drawing a line through the scatter. Its failure mode is known: the older people in the sample are the ones who survived and were willing to run a treadmill to exhaustion. It flatters age.

The correction came from following the same people over time. A 2005 analysis of the Baltimore Longitudinal Study of Aging took serial measurements of peak treadmill oxygen consumption in 375 women and 435 men aged 21 to 87, a community-dwelling cohort free of clinical heart disease, over a median follow-up of 7.9 years. Peak VO2 fell in all six age decades in both sexes, but the shape was the surprise: the rate of decline accelerated from 3% to 6% per 10 years in the 20s and 30s to more than 20% per 10 years in the 70s and beyond, and was larger in men than in women from the 40s onward. And the same pattern held in all quartiles of self-reported leisure-time physical activity, whether peak VO2 was expressed per kilogram of body weight or per kilogram of fat-free mass.

A second longitudinal cohort found the same architecture in a general population. A 2025 study of 427 Swedish adults born in 1958, assessed repeatedly from age 16 to 63, reported that estimated maximal aerobic capacity peaked at ages 26 to 36 in both sexes, then declined at 0.3% to 0.6% per year at first, accelerating to 2.0% to 2.5% per year, with no sex difference in decline rates; total decline from peak to age 63 across its physical measures ranged from 30% to 48%. Group variance also rose sharply — a 25-fold increase in the variance of relative aerobic capacity from adolescence to age 63. Different cohorts, different equipment, figures not interchangeable — but they agree on the shape: the curve bends downward, and the population fans out.

Why your maximum heart rate is the part you cannot negotiate with

How much oxygen you use per minute is two things multiplied together: how many times your heart beats, and how much oxygen your body actually takes up per beat — what the Baltimore analysis calls the oxygen pulse, the oxygen utilisation per heartbeat. The first term falls with age in a way largely indifferent to what you do.

The most-quoted formula, 220 minus age, was never validated in a sample with enough older adults. A 2001 re-analysis pooled group mean maximal heart rate values from 351 studies involving 492 groups and 18,712 subjects, then cross-validated in a laboratory study of 514 healthy subjects. The meta-analysis produced 208 minus 0.7 times age; the laboratory study, 209 minus 0.7 times age. The line was not different between men and women and was not influenced by wide variations in habitual physical activity levels. The authors concluded that the familiar equation underestimates maximal heart rate in older adults — so any zone computed as a percentage of it lands lower than intended: a bias toward training too easy, in the age group with the least margin.

The Baltimore data show how much of the decline that ceiling accounts for. Over that follow-up, the longitudinal rate of heart rate decline averaged only 4% to 6% per 10 years and accelerated only minimally with age, while peak VO2 fell faster and accelerated sharply. What did mirror the fall in peak VO2 was the oxygen used per heartbeat — what that analysis calls the oxygen pulse. So the accelerating part is not mostly the falling ceiling; it is the per-beat term. Whether that term can be moved is a question for the controlled trials below. The ceiling is not up for negotiation: the 2001 re-analysis found maximal heart rate unaffected by wide variations in habitual physical activity.

The athlete paradox: a higher line, not a flatter slope

This is the part of the literature most often reported backwards. The optimistic version comes from a 1990 follow-up of 15 well-trained master endurance athletes, mean age 62.0, and 14 sedentary controls, mean age 61.4, retested after roughly eight years. The sedentary men's VO2max fell by an average of 3.3 ml/kg/min, from 33.9 to 30.6 — 12% per decade — and their maximal heart rate fell 8 beats per minute, from 171 to 163. The athletes, who kept training, fell 2.2 ml/kg/min, from 54.0 to 51.8, a 5.5% decline per decade, with maximal heart rate unchanged at 171. A longitudinal design, which is a real strength — and a small, self-selected, non-randomised sample, the athletes already a decade into training before the first test.

The pessimistic version is larger. A cross-sectional laboratory study of 153 healthy men aged 20 to 75 — 64 sedentary, 89 endurance trained — found the absolute rate of decline in VO2max was greater in the trained men, -5.4 versus -3.9 ml/kg/min per decade, while the relative rate was similar. VO2max declined linearly across the age range in sedentary men but was maintained in trained men until approximately 50 years of age; the accelerated decline after 50 was related to falling training volume and associated with an increase in 10-km running time. And a meta-analysis of 242 studies covering 538 subject groups and 13,828 men found no difference between sedentary, active and endurance-trained groups in absolute rates of decline (-4.0, -4.0 and -4.6 ml/kg/min per decade) or relative rates (-8.7%, -7.3% and -6.8% per decade), and no difference in the rate of age-related reduction in maximal heart rate — contradicting the master-athlete follow-up on that point.

Two cautions. Both studies that disagree with the optimistic account are cross-sectional: they compare different people of different ages at one moment, and cannot tell you what training does to an individual over time. And all three are in men; the longitudinal evidence in both sexes found the acceleration persisting across activity quartiles.

So the supportable claim is narrower than “exercise slows aging” — and narrower than these designs can carry. Trained men are measured higher at every age: the meta-analysis of 242 studies found VO2max highest in endurance-trained and lowest in sedentary men, with active men in between; the two smaller studies compared only trained with sedentary men, and found the same ordering. What has not been shown here is that the slope flattens, or that someone who takes up training in their fifties moves onto the athletes’ line. Only the controlled trials speak to that, and they are the next section.

What a low number is actually associated with

The association between fitness and death in observational cohorts is unusually large, and needs unusually careful handling. The most striking dataset is a retrospective cohort of 122,007 patients referred for symptom-limited exercise treadmill testing at one academic medical centre, mean age 53.4, followed a median of 8.4 years, with 13,637 deaths across 1.1 million person-years. Risk-adjusted all-cause mortality was inversely proportional to fitness: elite performers versus low, adjusted hazard ratio 0.20 (95% CI 0.16-0.24). Inverted, low versus elite carried a hazard ratio of 5.04 (4.10-6.20) — larger than the same model's figures for coronary artery disease (1.29), smoking (1.41) or diabetes (1.40). There was no observed upper limit of benefit: elite versus high was 0.77 (0.63-0.95), and in patients 70 or older the elite-over-high advantage persisted at 0.71 (0.52-0.98).

A cohort of 750,302 US veterans aged 30 to 95, mean age 61.3, including 110,637 septuagenarians, 26,989 octogenarians and 45,232 women, followed a median of 10.2 years with 174,807 deaths, found the same graded inverse relationship across age, sex and race: lowest mortality risk at approximately 14.0 METs — hazard ratio 0.24 for men, 0.23 for women — and, for the least fit at the 20th percentile compared with extremely fit individuals, a roughly four-fold higher risk (4.09, 3.90-4.20); that is one tail measured against the other, not the unfit against the average. A meta-analysis of 33 cohort studies, with 102,980 participants and 6,910 deaths for all-cause mortality, put a number on the gradient: each 1-MET higher maximal aerobic capacity was associated with a pooled relative risk of 0.87 (0.84-0.90) for all-cause mortality and 0.85 (0.82-0.88) for coronary and cardiovascular events, one MET corresponding to about 1 km/h higher running or jogging speed. The American Heart Association's 2016 scientific statement argued on this basis for treating cardiorespiratory fitness as a clinical vital sign, on the grounds that it may be a stronger predictor of mortality than smoking, hypertension, high cholesterol or type 2 diabetes.

Now the discipline. Every cohort above is observational; none randomised anyone to be fit. Fitness is partly inherited, partly the residue of decades of behaviour, and partly a marker of not being ill — and that last one bites hardest in the treadmill-referral cohort, because those 122,007 people were referred for symptom-limited exercise testing, meaning a clinician had a reason to send them. Subclinical disease lowers treadmill performance before it kills you, and adjustment cannot remove what was never measured. A hazard ratio of 5.04 for being unfit is not a licence to say that becoming fit divides your risk by five.

The strongest evidence on change rather than status is a prospective study of 9,777 men examined twice, a mean of 4.9 years apart, then followed a mean of 5.1 years. Men unfit at both examinations had an age-adjusted all-cause death rate of 122.0 per 10,000 man-years; men fit at both, 39.6; men who improved from unfit to fit, 67.7 — a 44% reduction in mortality risk (95% CI 25% to 59%) relative to those who stayed unfit. Each one-minute increase in maximal treadmill time between examinations corresponded to a 7.9% decrease in mortality risk. Better than a snapshot, still not a randomised trial: the men who improved chose to, and people become fitter in midlife for reasons that include getting healthier. It is also, entirely, men.

One further association: a cohort of 370,980 dementia-free individuals, fitness estimated from a submaximal cycle test, followed a mean of 11.9 years, found high fitness associated with lower late-onset dementia risk both in those assessed under 55 (HR 0.58, 0.36-0.92) and over 55 (0.75, 0.63-0.89). Observational, estimated fitness, association.

What controlled training actually does to the number in older adults

A meta-analysis of 41 controlled clinical trials in 2,102 older subjects, within-group mean age 60 and over, found a pooled standardized effect size of 0.64, corresponding to a net increase in VO2max of 3.78 mL/kg/min (95% CI 3.24-4.33) versus control — a 16.3% improvement. Greater improvement was associated with training longer than 20 weeks and with an intensity of approximately 60% but less than 70% of VO2max.

A later dose-response meta-analysis, pooling 1,257 exercisers and 845 controls with a mean age of 67.45, reported a weighted net change of 3.78 ml/kg per min (95% CI 3.29-4.27). The identical central estimate should not be read as independent replication: both pool controlled endurance-training trials in older adults from the same literature, and their pooled samples are the same size (2,102 — here 1,257 exercisers plus 845 controls). The dose detail, though, is new, and it is the most actionable finding here. Differences between studies in VO2max change were significantly related to exercise intensity, which explained approximately 11% of the variance in response. VO2max improved significantly at 35% to 50% of heart rate reserve and kept improving as the dose rose; the largest adaptation came at a mean intensity of 66% to 73% of heart rate reserve, with the magnitudes at 57% to 65% and at 75% to 80% identical to one another. Intensities above 75% to 80% did not produce greater improvement and, the authors report, conversely resulted in large declines. Their optimum: 66% to 73% of heart rate reserve, 40 to 50 minutes per session, three to four days a week, for 30 to 40 weeks.

Roughly 3.78 mL/kg/min is not a transformation; the 41-trial meta-analysis that reported it describes the same net gain as a 16.3% improvement over control. Set beside the more than 20% per 10 years the Baltimore cohort recorded in the 70s and beyond, a one-off 16.3% is worth having and is not a reversal of aging — though the two figures come from different cohorts on different denominators, so this is a sense of scale, not a subtraction.

Dose matters in a graded way, and the cleanest demonstration is in a specific population. A randomised controlled trial in 464 sedentary, postmenopausal, overweight or obese women with elevated blood pressure — BMI 25.0 to 43.0, systolic pressure 120.0 to 159.9 mm Hg — assigned them to a non-exercise control group or to weekly energy expenditures of 4, 8 or 12 kcal/kg for six months, all training at the heart rate associated with 50% of each woman's peak VO2. Those doses worked out to means of 72.2, 135.8 and 191.7 minutes per week, and peak absolute oxygen consumption rose relative to control by 4.2%, 6.0% and 8.2%, with a significant trend across doses. Blood pressure did not change significantly in any exercise group versus control, despite elevated baseline pressures. Note who was studied before taking the percentages home: the transferable finding is the shape, not the figures.

Intervals come out ahead, by a margin smaller than the marketing

A systematic review of 14 studies in 429 middle-aged and older adults found VO2max improved within-group after both interval training (mean difference 2.26 mL/kg/min, 95% CI 1.50-3.02) and moderate-intensity continuous training (1.34, 0.45-2.23), with the gain significantly greater after interval training by a between-group mean difference of 1.10 mL/kg/min (0.55-1.64). A meta-analysis of randomised trials in older adults put the interval advantage in peak oxygen consumption at a weighted mean difference of 1.74 (0.80-2.69), and its subgroup analysis is the most concrete prescription in this literature: training periods longer than 12 weeks, two sessions per week, session lengths of 40 minutes, six sets and repetitions, more than 60 seconds per repetition, rest under 90 seconds.

A 2024 meta-analysis of 29 trials in 1,227 adults aged 60 and over complicates it usefully. Across all trials, interval and continuous moderate training produced significant and statistically similar effect sizes for VO2max (0.722 versus 0.490) — but restricted to controlled trials, the higher-quality subset, increases in VO2max occurred after interval training and not after continuous training (interval g = 1.068, p < 0.0001; continuous g = 0.109, p = 0.596).

So intervals come out ahead in these analyses — by 1.10 mL/kg/min in one and 1.74 in another — but all three reviews pool randomised comparisons of interval against continuous training in the same older-adult literature, so this is one body of evidence looked at three times, not three independent findings; and in the 2024 analysis of 29 trials the two modalities produced statistically similar effects until the analysis was restricted to controlled trials. A real advantage, and a small one beside the gap between doing something and doing nothing. If intervals are the reason you would not train, do not do intervals.

What five years of supervised training in Norwegian seventy-somethings actually showed

The longest of the trials here is the Generation 100 study, and it is routinely reported as something it is not.

The trial randomised 1,567 older adults aged 70 to 77, drawn from 6,966 individuals born between 1936 and 1942 in Trondheim, to five years of two sessions weekly of high-intensity interval training at about 90% of peak heart rate (n=400), moderate-intensity continuous training at about 70% of peak heart rate (n=387), or following the national physical activity guidelines (n=780, control). Mean age was 72.8, 790 were women, 87.5% reported overall good health and 80% medium or high physical activity at baseline — a healthy, already active population, which limits how far the result travels.

The primary outcome was all-cause mortality, and the primary comparison was null: all-cause mortality did not differ between the control group and the combined moderate-plus-interval group.

The rest is less conclusive. Against an observed control mortality of 4.7%, the interval group showed an absolute risk reduction of 1.7 percentage points, hazard ratio 0.63 (95% CI 0.33 to 1.20); the moderate-continuous group an absolute increased risk of 1.2 percentage points, 1.24 (0.73 to 2.10); interval versus moderate-continuous as reference, an absolute risk reduction of 2.9 percentage points, 0.51 (0.25 to 1.02). Every one of those confidence intervals includes 1.0. The trial's own conclusion is correspondingly restrained: no effect on all-cause mortality compared with recommended activity levels, though a lower mortality trend was observed after interval training compared with controls and with moderate training. A trend, in a pre-specified exploratory comparison, with an interval that does not exclude no effect.

One wrinkle any honest account must include: the controls chose to perform more of their activity as high-intensity interval training than the participants assigned to moderate continuous training did, so they achieved an exercise dose at an intensity between the two intervention groups. This was supervised training against unsupervised guideline-following in people already active — not exercise against inactivity.

A later report of secondary outcomes from the same trial — the same 1,567 participants, not a second study, and therefore not independent replication — examined sarcopenia components. Compared with control, grip strength declined less in the interval group at Year 1 (estimated difference 0.98 kg, p < 0.001) and Year 3 (0.03 kg, p = 0.016). At Year 3 the interval group increased gait speed by 0.01 m/s, an estimated difference versus control of 0.03 m/s (p = 0.016). By Year 5 all three groups had declined from baseline in gait speed (control -0.07, moderate -0.10, interval -0.04 m/s). Proportions meeting a clinically defined sarcopenia definition at baseline, Year 3 and Year 5 were 22%, 58% and 63% in control; 20%, 55% and 67% in moderate; 19%, 50% and 56% in interval — odds of developing it significantly lower for interval training than control at Year 3 (OR 0.51, p = 0.018) and Year 5 (OR 0.47, p = 0.009). No effects were found for moderate continuous training.

Read together: five years of supervised interval training in relatively healthy Norwegian seventy-somethings did not demonstrate a mortality benefit over national guidelines, but was associated with better preservation of some function markers — on Year 3 differences measured in hundredths of a unit, with everybody's gait speed still declining by Year 5.

The parts that do not respond, and the people the average hides

Every effect size above is a mean, and the spread around these means is wide.

A review of the HERITAGE Family Study and related work found that VO2max responses to standardised training programs have ranged from almost no gain up to a 100% increase in large groups of sedentary individuals. Age, sex and ethnic origin were not major determinants of those differences. Pretraining level mattered considerably for some traits, such as submaximal exercise heart rate and blood pressure, but only minorly for VO2max. What did track was family: significant familial aggregation of training-response phenotypes, implicating shared environment and genetic factors.

Quantifying non-response requires an untrained control group to separate real differences from measurement noise. One analysis does exactly that. A pooled analysis of 117 people who completed six weeks of very-low-volume sprint interval training — two 10-to-20-second all-out cycling sprints per session — alongside 40 no-intervention controls found a standard deviation of individual responses of 2.39 mL/kg/min, exceeding the technical, biological and random within-subject variability of the measurement. Against a smallest worthwhile change of 1.75 mL/kg/min, the likely non-response rate was 18% (21 of 117), while 49% (57 of 117) showed increases likely larger than that threshold. Crucially, those participants had a mean age of 30. Non-response is real and measurable; the rate in sixty-year-olds is not something these abstracts establish.

On sex the evidence is in tension. A meta-analysis of eight studies in 175 healthy untrained men and women, age-matched and given sex-matched doses of endurance training, found substantially larger increases in absolute VO2max in men (mean difference +191 ml/min, 95% CI 99 to 283, p < 0.001) and a greater effect on relative VO2max (+1.95 ml/kg/min, 0.76 to 3.15, p = 0.001), with no heterogeneity between studies. The HERITAGE review, on a different dataset answering a different question, concluded that sex was not a major determinant of interindividual variation. Both can be true — an average difference between groups is not the same thing as the drivers of person-to-person spread — but the pooled evidence does show a smaller mean gain in women than in men given a sex-matched dose, from eight studies and 175 healthy untrained men and women whose ages the review does not report. Whether that difference holds after 50 is not something those studies address.

What the heart itself does, and what that does not prove

This section is about structure and function — ventricular stiffness, blood ejected. These are not clinical outcomes: nothing below shows that changing them prevented an admission or a death.

The trial evidence is a randomised controlled trial in 61 healthy, sedentary, middle-aged participants, aged 53 ± 5, assigned to two years of supervised high-intensity training (n=34) or attention control (n=27), with right heart catheterisation, three-dimensional echocardiography and preload manipulation. Fifty-three completed; adherence was 88%. VO2max increased by 18% in the training group (29.0 to 34.4) with no increase in controls (29.5 to 28.7), and left ventricular stiffness fell in the trained group (stiffness constant 0.072 to 0.051, p = 0.0018) with no change in controls (0.0635 to 0.062, p = 0.83). End-diastolic volume rose while pulmonary capillary wedge pressure was unchanged. Two years, thirty-four people, healthy and sedentary and middle-aged — not heart failure patients. The authors' framing is that training may protect against future heart failure with preserved ejection fraction: a hypothesis about an outcome, not a measurement of one.

The dose question was addressed in a cross-sectional study of 102 healthy seniors stratified by lifelong training history into sedentary (fewer than 2 sessions a week), casual (2 to 3), committed (4 to 5) and competitive Masters athletes (6 to 7). Left ventricular distensibility was greater in committed exercisers (by 21%) and competitive athletes (by 36%) than in the sedentary; stiffness constants ran 0.062 sedentary, 0.079 casual, 0.055 committed, 0.035 competitive. The authors concluded that low doses of casual lifelong exercise do not prevent the decreased compliance of sedentary aging, whereas four to five sessions a week throughout adulthood prevent most of these age-related changes. That is the most demanding number here, and it comes from a cross-sectional comparison of people who chose their own exercise histories. It cannot establish that taking up four sessions a week in your late fifties would produce that heart.

The same limitation applies to every athlete-versus-sedentary contrast here. Nine male master athletes aged 64 versus nine sedentary healthy men aged 63 showed VO2max of 50.4 versus 29.6 mL/kg/min, higher peak-exercise stroke volume (132 ± 6 versus 111 ± 6) and a larger resting end-diastolic volume (153 versus 132 mL, p = 0.009). Nine lifelong endurance athletes aged 81 — cross-country skiers, including a former Olympic champion — versus six untrained men aged 82 showed relative VO2max of 38 ± 1 versus 21 ± 1 ml/kg/min, higher heart rate at peak (160 versus 146 beats per minute), and muscle oxidative enzymes 54% and 42% higher, with PGC-1α and Tfam messenger RNA 135% and 80% greater. Striking numbers — and a design that cannot separate the training from everything else distinguishing an Olympic champion from an average octogenarian.

The molecular layer is similarly limited. A study of 12 weeks of interval, resistance and combined training in young and old humans found all three enhanced insulin sensitivity and lean mass, but only interval and combined training improved aerobic capacity and skeletal muscle mitochondrial respiration; interval training produced a more robust increase in gene transcripts, particularly in older adults, and reversed many age-related differences in the muscle proteome. A mechanism for why intervals move aerobic capacity — not evidence of a longevity benefit.

A protocol the evidence will actually underwrite

Assembled only from the studies above, with the population each finding came from named, because the population is the whole game. None of it is individualised advice — every trial cited screened its participants.

For the movement-pattern side of this — how to build sessions your joints will tolerate — see our Movement Lab; the wider set of evidence-graded programs lives in the Protocols hub.

What is oversold

That training will stop the decline. The Baltimore cohort found the acceleration in every activity quartile. Whatever training does, arresting the age-related fall in peak aerobic capacity is not among the things it has been shown to do.

That fit people keep a young maximum heart rate. One small longitudinal follow-up found master athletes' maximal heart rate unchanged over eight years while sedentary men's fell 8 beats per minute. Two much larger analyses found no effect of activity status on maximal heart rate or its rate of decline. The larger evidence wins.

Sprint and interval protocols marketed to older adults. A trainability meta-analysis of interval training, reporting a mean VO2max increase of 0.51 L/min across 334 subjects, restricted itself by design to healthy sedentary or recreationally active humans under 45, and its inclusion criteria required at least 10 minutes of high-intensity work per session — so it does not speak to minimal-dose sprint protocols either. The pooled non-response analysis of very-low-volume sprint training studied participants with a mean age of 30. Neither result has been shown to hold at 60.

The octogenarian skier as a target. Comparisons like that one, and like the master-athlete cardiac studies, show what a highly selected group looks like beside a small comparison group — not what training would do to you.

Mechanism dressed as outcome. Reduced ventricular stiffness, larger end-diastolic volumes, higher citrate synthase, more mitochondrial protein synthesis, a partially reversed muscle proteome — all measured, all real, none of them a demonstrated reduction in any clinical event.

Two papers from one trial as two pieces of evidence. The Generation 100 mortality paper and the Generation 100 sarcopenia paper report the same 1,567 randomised Norwegians. Counting them twice is how a trend becomes a consensus.

How you would know it is working

The measurement you can act on and the measurement on your wrist are not the same object.

A 2025 systematic review of consumer wearables screened 252 records and included 13 studies. Most examined Garmin smartwatches paired with a chest-belt heart rate sensor and the Firstbeat algorithm; in every case the reference VO2max was measured in a laboratory with a graded treadmill test while the estimate came from submaximal outdoor runs. In seven studies the wearables proved valid or acceptable for VO2max estimation against the gold standard, and three demonstrated validity for lactate threshold estimation. Studies in untrained adults, recreational athletes and highly trained athletes were equally represented; usefulness in elite endurance sport was judged questionable.

Read the limits carefully, because they define what your drifting number means. That review is a qualitative synthesis, not a pooled estimate; in every included study the estimate came from submaximal outdoor running while the reference came from a laboratory treadmill test, and the reviewers note that using two or more submaximal runs as the index test could improve validity, though further examination is needed; and they explicitly call for longitudinal studies monitoring accuracy over months or even years, because nobody has established that the algorithms are stable over time. Your watch's estimate is a plausible one-off approximation of a lab value. Whether a small drift across eighteen months means anything is a question the evidence does not answer.

Two more durable ways to tell. The change-in-fitness cohort quantified improvement in units anyone can collect: each one-minute increase in maximal treadmill time between two examinations corresponded to a 7.9% decrease in mortality risk in those 9,777 men. And the cohort meta-analysis expressed its gradient per MET, one MET corresponding to about 1 km/h higher running or jogging speed. Neither is a promise about you. But a repeatable standardised effort — the same hill, the same treadmill protocol, timed — is closer to what those studies measured than anything an algorithm infers from a submaximal jog. Log it where you will still have it in a year; our Biomarker Tracker is built for that kind of slow series.

When the right answer is not to train harder

One caution, not drawn from the studies above, because those were trials and cohorts in screened participants rather than investigations of symptoms. Chest pain, pressure or tightness with exertion; fainting or near-fainting during or just after exercise; breathlessness that is new, disproportionate to the effort, or worse than it was a month ago; or a heart rhythm that feels wrong at rest — these are reasons to stop and speak to a clinician, not signals to push through. One of the large fitness-mortality cohorts here (122,007 patients) was assembled entirely from people referred for symptom-limited exercise testing: getting evaluated is what those people did, and it is the right order of operations.

Which brings us back to the watch, and the hill. The evidence will not let anyone tell you that a training block returns you to the number you had at forty, and it will not let anyone tell you the drift is meaningless. What it supports is narrower: the descent is real and it steepens; controlled trials in older adults raise measured aerobic capacity by a few millilitres per kilogram per minute with months of consistent moderate work; how much any one person gains varies enormously, from almost no gain up to a 100% increase in large groups of sedentary individuals, with familial factors rather than age or sex tracking that spread; where intervals have been compared head-to-head with continuous training in this age group they came out ahead by about 1 to 2 mL/kg/min; and in the study that measured change rather than status, the men who became fit did better than the men who stayed unfit. A case for training, not for certainty — and the hill will still be a hill.

Educational, not medical advice.