Estimated Cardiorespiratory Fitness Was Linked to Lower Disease Risk in 3 Large Cohorts

A new analysis of more than half a million adults asked whether cardiorespiratory fitness estimated from ordinary survey data can stand in for a treadmill test. The shortcut held up in Britain and the United States, and did not hold in China.

Cardiorespiratory Fitness is among the strongest single markers researchers have for who stays healthy and who does not, yet measuring it properly requires a graded exercise test that most population studies cannot afford to run. A study published on September 25, 2026 in BMC Cardiovascular Disorders put the practical alternative on trial: equations that estimate fitness from variables a questionnaire can already collect.

The work was led from the Nanshan School at Guangzhou Medical University and the First Affiliated Hospital of Guangzhou Medical University, with Shuben Li as corresponding author. The team applied validated non-exercise equations to three of the largest ageing cohorts assembled anywhere: the UK Biobank in Britain, the Health and Retirement Study in the United States, and the China Health and Retirement Longitudinal Study.

Participants were sorted into quartiles of estimated cardiorespiratory fitness, and Cox proportional hazards models produced multivariable-adjusted hazard ratios with 95% confidence intervals. In the UK Biobank the outcomes covered all-cause mortality, cardiovascular mortality, cancer mortality, incident cardiovascular disease, incident diabetes and incident cancer. In the two retirement cohorts, all-cause mortality and incident cardiovascular disease were assessed.

In the British and American cohorts the pattern was consistent. Higher estimated cardiorespiratory fitness went with lower risk of death and of major chronic disease, and the association was, in the authors words, particularly pronounced for incident diabetes. In the Chinese cohort the signal was absent.

Why Estimated Cardiorespiratory Fitness Is Worth the Trouble

The reference method is a graded exercise test with gas analysis, which yields a maximal oxygen uptake value. It is accurate. It is also slow, expensive and physically demanding, and it cannot realistically be run on half a million people scattered across three continents.

Non-exercise equations sidestep the treadmill entirely. They combine variables that large surveys already capture and return a fitness estimate on a familiar scale. The appeal is obvious. The open question, as the paper frames it, is whether estimated cardiorespiratory fitness carries the same predictive weight as a measured value across diverse populations.

That question is not academic. Risk models, screening tools and public health targets increasingly lean on fitness as an input, and if the input is a proxy, every conclusion built on it inherits the proxys error. A cheap estimate that works is enormously valuable. A cheap estimate that quietly fails in some populations is worse than none, because it looks like evidence.

Readers who followed our report on a framework that standardized VO2 max across five different exercise tests will recognize the underlying problem. Fitness is easy to define and stubbornly hard to compare from one study or one method to the next.

What Each of the Three Cohorts Was Asked to Show

The UK Biobank carried the heaviest analytical load, supplying six separate outcomes spanning death, cardiovascular disease, diabetes and cancer. That breadth is what allows a single study to say something about general health rather than about one disease.

The Health and Retirement Study and the China Health and Retirement Longitudinal Study contributed two outcomes each, all-cause mortality and incident cardiovascular disease. Both are long-running surveys of older adults, built for the same broad purpose in very different places.

That geographic spread was the point. Our earlier coverage of exercise participation and relative health in a survey of 76,827 Chinese adults showed how differently activity patterns can register from one national sample to another.

The Diabetes Signal Stood Out From the Rest

Of every outcome examined, incident diabetes showed the most pronounced association with estimated cardiorespiratory fitness. The authors single it out in the abstract, and it is the finding most likely to shape how this tool gets used.

That emphasis is worth sitting with. Diabetes risk is modifiable, it is screened for routinely, and a cheap survey-based marker that flags it early would be genuinely useful in primary care and in public health planning.

It is also the outcome where a fitness estimate has the most room to be informative. Cardiovascular events and cancer are shaped by a long list of factors that surveys capture poorly, while the metabolic picture that cardiorespiratory fitness reflects sits closer to the mechanism that drives diabetes risk in the first place.

Where the Cardiorespiratory Fitness Equations Appear to Break Down

In the China Health and Retirement Longitudinal Study, estimated cardiorespiratory fitness was not significantly associated with all-cause mortality or with incident cardiovascular disease. Not merely a weaker association, on the abstracts account, but not a statistically significant one.

The authors do not read this as evidence that fitness stops mattering east of the Urals. Their stated interpretation is narrower and more technical: current equations may require population-specific calibration, and further validation in Asian populations is needed before the estimates can be trusted there.

That distinction matters. The equations were developed and validated largely in Western samples. If the relationships between the inputs and true measured cardiorespiratory fitness differ by population, the output drifts, and a drifting estimate predicts poorly no matter how real the underlying biology is.

Straight Lines, Curves and What the Splines Suggested

Beyond the quartile comparisons, the team fitted restricted cubic spline models to look at shape rather than category. In the UK Biobank the associations were nonlinear and inverse. In the other two cohorts the patterns were approximately linear.

A nonlinear inverse curve usually means the steepest benefit sits at the bottom of the range, where the least fit people are, and flattens as fitness climbs. It is the same shape that has turned up repeatedly in activity research, including our report on 12 minutes of vigorous activity a week matching 150 moderate minutes for lower mortality.

What the Subgroup Analyses Added to the Picture

The team also ran subgroup analyses to test whether the associations held within slices of each cohort rather than only in the whole. According to the abstract, those analyses generally supported consistent protective associations in the European and North American samples.

Consistency across subgroups is a modest but meaningful form of reassurance. It suggests the relationship between estimated cardiorespiratory fitness and later health was not being carried by one unusual slice of the data, which is a common way for large-cohort findings to look stronger than they are.

It also sharpens the contrast with the Chinese cohort. A result that survives subgroup slicing in two populations and fails to appear at all in a third points more toward a measurement problem than toward genuine biological difference, which is precisely the reading the authors offer.

What Cardiorespiratory Fitness Findings Mean for Everyday Training

Nothing in this study tells any individual reader what to do on Tuesday morning. It is a methods paper as much as a health paper, and its audience is researchers deciding whether to trust a shortcut.

Still, a few things follow reasonably from it. Cardiorespiratory fitness keeps earning its place as a health marker rather than a vanity metric, and the outcomes it tracked here are the ones people actually care about avoiding.

It is also a reminder to hold estimates loosely. A number that arrives without a mask, a treadmill and a technician is an inference, and the quality of that inference depends on how closely you resemble the people the equation was built on.

• Treat any cardiorespiratory fitness number produced without an exercise test as an estimate, including the readouts on consumer wearables.
• Do not assume a risk equation developed in one population transfers cleanly to another.
• The diabetes finding is an association in observational data, not a promise that training prevents diabetes.
• If you want a real number, a supervised exercise test remains the reference method.

For the training side of that equation, our coverage of structured exercise and survival in people with cancer gives a sense of what supervised programs have shown in clinical populations, where the evidence is built on trials rather than surveys.