What VO₂max measures — and what it does not
VO₂max is the maximal rate at which the body can take up, transport and use oxygen, conventionally expressed in millilitres of oxygen per kilogram of bodyweight per minute. It integrates the whole chain — ventilation, cardiac output, the oxygen-carrying capacity of the blood, capillary delivery and mitochondrial extraction — into a single number.
The reference review of what limits it concludes that in healthy people at sea level the constraint is mainly delivery: maximal cardiac output, driven largely by stroke volume, rather than the muscle’s capacity to consume oxygen. Pulmonary diffusion, blood oxygen content and muscle capillarisation all contribute, and their relative weight shifts at altitude and in highly trained athletes.
That same review separates VO₂max from endurance performance, which it treats as the product of three things: the maximal ceiling, the fraction of it that can be sustained, and exercise economy. VO₂max sets an upper bound. It does not by itself determine what happens over a race.
The mortality association, at scale
Cardiorespiratory fitness is among the strongest predictors of mortality in observational medicine. The 2009 JAMA meta-analysis quantified it in healthy men and women: each one-MET higher fitness — one metabolic equivalent, the standard unit of exercise intensity — was associated with roughly 13% lower all-cause mortality and roughly 15% lower risk of coronary heart disease and cardiovascular events.
Two later syntheses extend it. A 2022 meta-analysis of 37 cohort studies and 2,258,029 participants restricted itself to fitness measured objectively by exercise testing rather than estimated, and found the same inverse, graded relationship with all-cause mortality. A 2024 overview of meta-analyses covering 199 unique cohorts and more than 20.9 million observations reports the association as strong and consistent across morbidity and mortality outcomes alike.
All of it is observational. These studies measure fitness in people who already differ in income, smoking, diet, existing illness and much else besides, then follow them; adjustment narrows those differences but cannot remove them. The size of the evidence base makes this about as certain as an association gets. It does not convert it into a causal claim.
What a change in fitness tracks
Baseline fitness may be prognostic largely because it summarises everything that came before the measurement. The more informative design measures a change. The Ball State Adult Fitness Longitudinal Lifestyle Study assessed VO₂max directly, ran participants through a short-term supervised training period, reassessed them, and then followed mortality.
Participants whose measured fitness improved carried lower subsequent mortality risk than those whose fitness did not. Because the exposure is a within-person change occurring after enrolment, it is harder to dismiss as a pure marker of pre-existing constitution than a single baseline reading is.
It remains observational. Who responds to a training period and who does not is not random, and the same factors that predict a fitness improvement — adherence, health status, age, inherited response — also predict survival. The finding is that a change in fitness is prognostic. It is not proof that producing the change produces the outcome.
Fitness against fatness
Fitness and body size are correlated, and both are associated with mortality, which leaves open the question of which the risk actually attaches to. A meta-analysis addressed it by classifying participants on both at once and comparing each group against fit, normal-weight participants.
Unfit participants carried roughly double the all-cause mortality risk of the fit reference group regardless of which adiposity category they fell into. Participants who were fit but classified as overweight or obese showed no significantly elevated risk relative to fit normal-weight participants.
A measurement asymmetry matters when reading that. Fitness assessed by exercise testing is a far more precise variable than BMI is a measure of body fatness, and in a joint model the better-measured of two correlated exposures tends to absorb the association they share. The conclusion that fitness carries prognostic information independent of body size is solid. The stronger reading — that body size does not matter — is not what this design can support.
Intervals against continuous work
The comparison dominating this literature is interval training against continuous endurance training. A meta-analysis of controlled trials found both raised VO₂max significantly, with the pooled estimate favouring interval training by a modest margin. A separate synthesis of sprint-interval training — protocols built around repeated all-out efforts of roughly thirty seconds — reports improvements in VO₂max and in aerobic exercise performance among untrained and recreationally trained participants.
A further review moved the outcome beyond fitness. Examining high-intensity interval training against cardiometabolic health, it reported improvements in body composition, blood pressure and markers of glucose handling, with the largest effects in participants who were overweight or in poorer metabolic health at the start.
The qualifications sit in the designs. Most of these trials run for weeks to a few months, in supervised settings, in participants with substantial room to improve — conditions that favour whichever stimulus produces the quickest early change and say little about what separates the two over years. Those trial conditions also say little about what any one person would get in ordinary life.
The average is not the individual
Every pooled estimate above is a group mean. A meta-analysis of VO₂max trainability reports how widely responses to high-intensity interval protocols vary — between studies in their mean improvement, and more importantly between individuals within them.
That distribution includes participants who complete an identical programme and show little measurable change in maximal oxygen uptake. This is not a marginal caveat to the averages; it is a property of them. A mean improvement describes what happened to a group, and the range of individual outcomes behind any such mean is far wider than the mean alone conveys.
Some of the spread is measurement rather than biology — VO₂max testing carries its own error, and short trials are exposed to regression to the mean. Some of it is dose. The review of how exercise intensity is prescribed shows that the standard anchors, whether a percentage of maximum heart rate, of heart-rate reserve or of VO₂max, place different individuals at different points relative to their own physiological thresholds, so the same nominal intensity is not the same stimulus. Between measurement error and unequal dosing, a headline average is a weaker guide to any one person than its precision implies.
