Your Aerobic Capacity Falls Twice As Fast As The Textbooks Say

RapAdmin · September 6, 2026, 5:02am

Three Norwegian exercise physiologists compared how fast VO2peak declines with age when measured two different ways: by testing many people of different ages once (cross-sectional), versus testing the same people repeatedly over years (longitudinal). The cross-sectional literature, which is what almost every clinical reference chart is built from, produces the familiar figure of roughly 10 percent lost per decade, holding steady across the lifespan. The longitudinal literature says something different and worse. Decline is not linear. It accelerates. After age 70 it reaches 15 to 20 percent per decade in women and 20 to 25 percent per decade in men. The gap between the two designs is largest exactly where it matters most clinically, in the oldest patients, and the likely cause is survivor bias: the 80-year-olds who show up for a maximal treadmill test were never average 40-year-olds.

There is a number that shows up in nearly every discussion of aerobic fitness and aging: 10 percent per decade. It is tidy, it is memorable, and according to this review it is wrong in the age range where it gets used most.

The problem is architectural. To build a fitness reference chart you need thousands of people, so researchers recruit a wide age range, test everyone once, and draw a line through the results. That line is a snapshot of different people, not a trajectory of one person. It compares today’s 30-year-olds to today’s 80-year-olds, and the 80-year-olds in that room are a survivor population. Low fitness predicts death and disease with unusual reliability, so the people who were unfit at 50 are disproportionately absent from the 80-year-old group. What looks like a gentle slope is partly a filter.

Follow the same individuals instead and the slope steepens. Two large longitudinal cohorts, the Baltimore Longitudinal Study of Aging in the United States and the HUNT Study in Norway, tracked participants across a wide age range and reported results by decade. Both found acceleration. Both found the acceleration is worse in men. In men past 70, aerobic capacity falls by roughly a quarter every ten years.

The designs disagree in a specific, age-dependent pattern. In young adults, cross-sectional estimates run slightly high. Through middle age the two converge. After about 60 the longitudinal estimates pull decisively ahead. One repeated-measures study of older adults put the annual loss at 0.69 mL/kg/min per year in men when tracked over time, against 0.34 when estimated across ages, a factor of two.

The clinical consequence is direct. A physician following a patient’s fitness over a decade, and comparing it against a cross-sectional chart, will see a normal-looking decline and reassure. The patient may in fact be losing capacity at the expected longitudinal rate, which is far larger, or at an abnormal rate hidden inside a too-generous expectation. Trajectory needs a trajectory-derived reference.

The review also leaves a hard finding sitting in plain sight. In the Generation 100 trial, 1567 adults over 70 were followed for five years, and VO2peak fell about 2 percent per year after the first year in the supervised exercise groups and the control group alike, despite training volume being maintained. Twenty percent per decade, with the exercise already priced in.

Actionable Insights

Four things follow from this paper.

First, your own previous test result is a better benchmark than any age-matched chart, particularly after 60. Charts built from one-time testing of many people understate how fast a real person declines.

Second, expect acceleration. Planning around a steady 10 percent per decade will overestimate where you land at 80. A man testing with a VO2Max at 31 mL/kg/min at age 70 would project to about 25 by age 90 under the 10 percent rule, but to roughly 17 under the observed 25 percent rule. That difference of about 8 mL/kg/min is more than 2 METs, and 18 mL/kg/min is near the level where daily independent living becomes effortful.

Third, the size of a fitness difference is worth knowing in risk terms. One MET, which is 3.5 mL/kg/min, is associated with roughly 13 percent lower all-cause mortality and 15 percent lower cardiovascular event risk in meta-analysis. The measurement-design gap alone, about 3.5 mL/kg/min per decade in older men, is one whole MET.

Fourth, and least comfortable: in the one large randomized trial cited, five years of supervised training in over-70s produced a VO2peak advantage over controls of 0.76 mL/kg/min. That is a small effect, worth about a year of delay in decline. Peak aerobic reserve built before 60 appears to matter more than training started after 70. Build the VO2Max ceiling early.

Context and Source

Open Access Paper: Age-related decline in peak oxygen uptake: Cross-sectional vs. longitudinal findings. A review., 2023 Jan 13.
Institution: Department of Circulation and Medical Imaging, Norwegian University of Science and Technology (NTNU), Trondheim, with St. Olavs University Hospital, Norway, and the University of Queensland, Australia.
Country: Norway (with Australian affiliation).
Journal: International Journal of Cardiology Cardiovascular Risk and Prevention.
Journal impact evaluation: This journal is a companion title to International Journal of Cardiology and does not carry a confirmed Clarivate Journal Impact Factor in the JCR core collection.

Biomarker Data (Effect Size Extraction)

Reported decline rates

The primary outcome throughout is one biomarker: VO2peak, the maximum oxygen a person can consume during exercise to exhaustion. It is measured by gas analysis, which is the gold standard rather than an estimate from a watch or a step test. It is also one of the strongest single predictors of death in medicine, which is why the size of its decline matters.

Median and maximum values, mapped onto what this paper actually reports

Since there is no lifespan outcome, the closest honest translation of “median and maximum improvement” is the typical versus the worst-case rate of decline reported across the literature reviewed.

Median or typical decline, from cross-sectional studies (the one-time testing designs that clinical reference charts are built from):

0.40 mL/kg/min lost per year in men, from meta-analysis, the same in active and sedentary men.
0.44 in physically active women and 0.35 in sedentary women, from meta-analysis.
Published prediction equations across many countries cluster between 0.30 and 0.50 per year.
Expressed as a percentage: about 10 percent per decade, assumed constant across the lifespan.

Maximum or worst-case decline, from longitudinal studies (the same individuals retested years later):

Men over 70: 20 to 25 percent per decade, found independently in both large cohorts, the Baltimore Longitudinal Study of Aging in the United States and the HUNT Study in Norway.
Women over 70: 15 to 20 percent per decade.
Fastest single reported annual rate: 0.69 mL/kg/min per year in older men (Hollenberg, 592 adults with a mean age of 65, repeated testing over 6.3 years).

The spread between the typical figure and the worst-case figure is the entire point of the paper. In percentage terms it is 10 versus 25 per decade, a factor of 2.5. [Confidence: High on the direction and that the gap is real; Medium on the exact endpoints, since both come from two cohorts with 8 to 10 year observation windows extrapolated out to a decade]

Effect size 1: the exposure itself, fitness versus death risk

This is the anchor that makes every other number below interpretable. It comes from an external source, the Kodama 2009 meta-analysis in JAMA, not from this review.

Per 1 MET higher fitness (3.5 mL/kg/min): relative risk of death from any cause 0.87, meaning 13 percent lower risk.
Per 1 MET higher fitness: relative risk of coronary or cardiovascular events 0.85, meaning 15 percent lower risk.
Per 2 METs (7 mL/kg/min): 0.87 squared equals 0.76, meaning 24 percent lower risk of death.

Plain reading: a difference of about 3.5 mL/kg/min between two otherwise similar people is associated with roughly one in eight fewer deaths over follow-up. Note this is a comparison between different people, not proof that raising your own number by 1 MET delivers the same benefit. [Confidence: High on the association, Medium on applying it to within-person change]

Effect size 2: how much the two study designs disagree

From Hollenberg, the one study in the review reporting both estimates in the same people. Annual decline, longitudinal versus cross-sectional:

Men, 0.69 versus 0.34 mL/kg/min per year.

Ratio: 2.03, so the longitudinal decline is twice the cross-sectional one.
Absolute gap over ten years: 3.5 mL/kg/min, which is exactly 1.0 MET.
As a percentage of a typical 70-year-old man’s 31 mL/kg/min: 11 percent of his total capacity, per decade, unaccounted for.
Cohen’s d, using an assumed male standard deviation of 6 mL/kg/min: 3.5 divided by 6 equals 0.58, a moderate effect.
Translated through Kodama: 1 MET equals a risk ratio of 0.87, so the reference-chart error hides about 13 percent of mortality-relevant change per decade.

Women, 0.39 versus 0.23 mL/kg/min per year.

Ratio: 1.70.
Absolute gap over ten years: 1.6 mL/kg/min, which is 0.46 MET.
As a percentage of a typical 70-year-old woman’s 26 mL/kg/min: 6 percent per decade.
Cohen’s d, using an assumed female standard deviation of 5 mL/kg/min: 1.6 divided by 5 equals 0.32, small to moderate.
Translated through Kodama: about 6 percent of mortality-relevant change hidden per decade.

Plain reading: judging an older man’s ten-year fitness trajectory against a standard age chart quietly forgives him a full MET of loss. [Confidence: High on the arithmetic, Medium on the mortality translation]

Effect size 3: how big one decade of aging actually is

Using the review’s own longitudinal figures and typical starting values.

Man aged 70, starting at 31 mL/kg/min, declining 25 percent per decade:

Absolute loss: 7.75 mL/kg/min in ten years.
In METs: 2.21 METs.
Cohen’s d, assumed SD 6: 1.29. Large. This is the size of change that moves a man out of his fitness percentile band entirely.
Risk translation: losing 2.21 METs corresponds to a risk multiplier of about 1.36, so roughly 36 percent higher all-cause mortality risk than his 70-year-old self, from the fitness change alone.

Woman aged 70, starting at 26 mL/kg/min, declining 17.5 percent per decade:

Absolute loss: 4.55 mL/kg/min in ten years.
In METs: 1.30 METs.
Cohen’s d, assumed SD 5: 0.91. Large.
Risk translation: risk multiplier about 1.20, so roughly 20 percent higher mortality risk.

Compare against what the conventional 10 percent per decade rule predicts for the same man: 3.1 mL/kg/min lost, 0.89 MET, Cohen’s d of 0.52, moderate rather than large. The heuristic and the observation differ by three quarters of a population standard deviation per decade. [Confidence: Low to Medium, because both standard deviations are imported from external reference data rather than reported in this paper. The direction is robust, the decimal places are not.]

Effect size 4: the actual treatment effect, with both arms reported

This is the only randomized treatment-versus-control comparison anywhere in the review’s evidence base: the Generation 100 trial, 1567 Norwegian adults over 70, five years.

Groups and baselines:

Control, n = 780, advised to follow national physical activity guidelines. Baseline VO2peak 31 mL/kg/min in men, 26 in women.
Moderate continuous training, n = 387. Baseline 32 in men, 26 in women.
High intensity interval training, n = 400. Baseline 31 in men, 26 in women.

Outcome after five years:

HIIT versus control: 0.76 mL/kg/min higher (99 percent confidence interval 0.02 to 1.51).
Both exercise arms combined versus control: 0.39 mL/kg/min higher (99 percent CI minus 0.22 to 1.00), which includes zero and is therefore not distinguishable from no effect.
Moderate continuous training versus control: no significant difference.
All groups, control included, lost about 2 percent per year after the first year, equivalent to 20 percent per decade, despite exercise volume being maintained.

Effect size of the best-performing arm, HIIT versus control:

Absolute: 0.76 mL/kg/min.
Percentage: 2.5 percent of the 31 mL/kg/min starting value.
In METs: 0.22 MET.
Cohen’s d, assumed SD 6: 0.76 divided by 6 equals 0.13. Small, well below the 0.2 threshold at which an effect is normally considered even modest.
Risk translation through Kodama: about 3 percent lower all-cause mortality risk. The confidence interval spans 0.1 percent to 5.8 percent, so the honest range runs from almost nothing to modest.
Time translation: at a decline of roughly 0.62 mL/kg/min per year in this age group, an advantage of 0.76 buys about 1.2 years of delay, bought with five years of supervised training.

Plain reading: five years of supervised high intensity training in adults over 70 produced a benefit around one tenth the size of a decade of aging in the same population. The lower bound of the confidence interval is 0.02, effectively zero, so even that small benefit is not securely established. This is the single most consequential number in the review for anyone building a protocol, and the review mentions it in one sentence. [Confidence: High that the effect is small; Medium on interpretation, because the control group was actively counselled on activity guidelines and many controls trained hard, compressing the contrast between arms]

Summary table of effect sizes

What is being measured	Absolute size	Percentage	Standardized effect
One decade of aging, man over 70	7.75 mL/kg/min (2.21 METs)	25 percent of baseline	Cohen’s d 1.29; risk multiplier 1.36
One decade of aging, woman over 70	4.55 mL/kg/min (1.30 METs)	17.5 percent of baseline	Cohen’s d 0.91; risk multiplier 1.20
Same decade under the 10 percent rule, man	3.10 mL/kg/min (0.89 MET)	10 percent of baseline	Cohen’s d 0.52; risk multiplier 1.13
Cross-sectional versus longitudinal gap, men	3.5 mL/kg/min per decade (1.0 MET)	11 percent of baseline	Cohen’s d 0.58; 13 percent of risk hidden
Cross-sectional versus longitudinal gap, women	1.6 mL/kg/min per decade (0.46 MET)	6 percent of baseline	Cohen’s d 0.32; 6 percent of risk hidden
Five years of supervised HIIT versus control, over-70s	0.76 mL/kg/min (0.22 MET)	2.5 percent of baseline	Cohen’s d 0.13; risk multiplier 0.97

Reading the table in one line: aging is a large effect, the measurement error in how we track it is a moderate effect, and the best available exercise intervention in the elderly is a small effect. All Cohen’s d values assume external standard deviations of 6 mL/kg/min in men and 5 in women, because this paper reports none. All risk multipliers apply an external per-MET mortality estimate.

Secular trend, for context

Pooled data across eight high and upper-middle income countries showed cardiorespiratory fitness declining about 1.6 percent per decade at population level between the 1960s and 2016. That is roughly one sixth the size of the individual aging effect, and it pushes in the same direction for cross-sectional comparisons, since people born earlier were fitter at any given age than people born later. It contributes to the design discrepancy without explaining it.

The phenotype in physiological terms

VO2peak is the product of cardiac output (heart rate times stroke volume) and arteriovenous oxygen difference (peripheral extraction). Any accelerating decline must be driven by acceleration in one of those terms.

Maximum heart rate declines close to linearly, roughly 0.7 beats per year. That is a linear input and cannot by itself produce a non-linear output. Stroke volume at peak effort is relatively preserved into old age in the healthy. This points to the periphery, specifically the arteriovenous oxygen difference, as the accelerating term: skeletal muscle mass, capillary density, and mitochondrial oxidative capacity. [Confidence: Medium]

Organ-specific aging priorities implied

Priority one is skeletal muscle, not the heart. Sarcopenic loss accelerates after 70, and VO2peak expressed per kilogram of body mass falls both because the numerator falls (less oxidative muscle) and because the denominator holds or rises (fat mass replacing lean mass). The review does not decompose this, which is a real weakness, but it is the most parsimonious explanation for the sex difference the authors report. Men lose more absolute lean mass and their peak capacity was higher, so they have more to lose and lose it faster.

Priority two is mitochondrial quality control in muscle. Priority three is the vasculature, specifically arterial stiffening and reduced exercise hyperemia limiting oxygen delivery. Priority four is central cardiac function, which appears to be the least of the four in healthy aging.

Rapamycin Longevity News