Adding treadmill fitness to standard risk factors sharpened 10-year death-risk estimates in 562,234 adults

Estimates of a person’s risk of dying over the next decade usually rest on a short list of facts: age, sex, blood pressure, diabetes, cholesterol and smoking. How far someone can go on a treadmill is not on that list, and a study of more than half a million adults suggests the estimates get better when it is added.

In the study, published on October 6 in Medicine & Science in Sports & Exercise, the least-fit fifth of participants had more than twice the 10-year risk of death of everyone else once the standard factors were accounted for. The gain in overall predictive accuracy was modest, though, and all of the participants were people who had been given a treadmill test as part of their medical care.

Why fitness and risk prediction is a separate question

Cardiorespiratory fitness (CRF) is the capacity of the heart, lungs and muscles to deliver and use oxygen during sustained effort. That fitter people live longer has been shown many times: a pooled analysis from 2009 combined 33 studies with 102,980 participants and found that each one-unit step up in fitness went with a 13% lower risk of death from any cause.

That unit is the metabolic equivalent, or MET, the energy the body uses at rest, so a person who peaks at 8 METs on a treadmill can work at eight times their resting rate. The 2009 analysis pictured one unit as corresponding to 1-km/h higher running or jogging speed.

Knowing that fitness tracks survival is not the same as knowing it helps predict survival for an individual. Age, blood pressure and diabetes already carry much of the same information, since unfit people are more often older and sicker. The useful question for a clinician is whether a fitness measurement moves anyone’s estimated risk once those factors are already in the calculation.

The American Heart Association took up that question in 2016, in a scientific statement subtitled “A Case for Fitness as a Clinical Vital Sign”, which set out to raise awareness of the added value of CRF for improving risk prediction.

The authors of the new paper say the matter was still open. Fitness is a powerful predictor of death, they write, but how much it adds beyond traditional cardiovascular risk factors when doctors sort patients by risk is still not fully pinned down.

How the fitness and mortality study worked

The research was led by Jonathan Myers and Peter Kokkinos, who also wrote the 2022 study of veterans described below. Their data came from a clinical exercise testing cohort, a group of people who had taken treadmill tests as part of their care and were then followed over time, and 562,234 adults with complete records were analyzed.

Each person’s fitness was the peak METs achieved during treadmill testing, and the researchers used it to split people in two. The least-fit fifth, those who peaked below 5.0 METs, were classed as unfit, and the other four fifths as fit.

Next they built two statistical models, formulas that turn a person’s characteristics into a probability, to estimate each person’s risk of dying from any cause within 10 years. The first used six standard inputs: age, sex, hypertension, diabetes, dyslipidemia, and smoking, where hypertension is high blood pressure and dyslipidemia is an unhealthy level of cholesterol or other blood fats. The second used the same six and added fitness.

People were followed for an average of 10.2 years, and the two models were then compared using three standard yardsticks for prediction.

What the study found about fitness and risk of death

As expected, fitness was linked to death independently of the standard factors. With those six held equal, each additional MET went with a 16% lower hazard of death, meaning deaths occurred at a 16% lower rate over time. With so many people in the study, the margin of uncertainty around that estimate was tight, running from 15.8% to 16.2%.

The comparison of the two groups gives the clearest picture. With the standard factors held equal, risk was substantially higher among unfit people: the estimated chance of dying within 10 years was 32.9% for the unfit group and 13.4% for the fit group.

Put another way, among people of the same age and sex with the same blood pressure, diabetes, cholesterol and smoking status, about one in three of the least fit would be expected to die within a decade, against about one in seven or eight of the rest.

What the prediction measures show about adding fitness

The three yardsticks each answer a slightly different question.

Ranking. The C-statistic measures how well a model ranks people. It is the probability that, of two people picked at random, one who died and one who survived, the model gave the higher risk to the one who died. A value of 0.5 is no better than a coin toss and 1.0 is perfect. The standard model scored 0.662. With fitness added it scored 0.682, a gain of 0.020.

Moving between categories. Net reclassification improvement asks whether the new model moves people in the right direction: those who went on to die into higher risk categories, and survivors into lower ones. Wrong moves are subtracted from right ones. For risk categories the figure was 0.06. A version that counts any shift in estimated risk, however small, gave 0.37.

Separation. A third measure, called integrated discrimination improvement, looks at how far apart the model places people who died and people who survived. It also improved.

All three moved in the same direction, and each gain passed the standard statistical test against chance. Their size is another matter. A rise from 0.662 to 0.682 leaves the model a long way from perfect, and a category figure of 0.06 is small. The authors read the results as showing that fitness meaningfully enhances individual-level risk stratification, the sorting of people into groups by level of risk. They argue for building fitness into routine clinical risk assessment frameworks.

Both things can be true together. A measure can mark out a group at much higher risk, as the 32.9% and 13.4% figures show, and still add only a little to how well a model ranks a whole population, because age already does so much of the ranking.

How this fits with earlier research on fitness and mortality

The same question was asked in 2011, on a smaller scale and for a narrower outcome, by researchers using the Cooper Center Longitudinal Study, who analyzed 66,371 people and 1,621 deaths from cardiovascular disease. They too started from the gap, noting that how much fitness improves risk classification on top of traditional risk factors was unclear. Adding fitness changed the risk category of 10.7% of the men, and the 10-year net reclassification improvement was 0.121. That is about twice the category figure in the new study, though the two are not directly comparable: one predicted cardiovascular death and the other death from any cause.

On the basic link between fitness and survival, the evidence is broad. At the Cleveland Clinic, 122,007 patients sent for a treadmill test were followed for a median of 8.4 years, during which 13,637 died. The fittest group had an 80% lower risk of death than the least fit, with no observed upper limit of benefit. The authors also reported that the extra risk tied to low fitness was comparable to or greater than that of traditional clinical risk factors such as smoking and diabetes.

Kokkinos, Myers and colleagues reported a similar pattern in 2022 among 750,302 US veterans, of whom 174,807 died, and the least-fit fifth had about four times the risk of the fittest 2%. Being unfit, the authors concluded, carried more risk than any of the cardiac risk factors examined.

A 2024 overview pulled the field together, covering 26 systematic reviews representing 199 cohort studies, and found each additional MET went with 11% to 17% lower mortality, a range that brackets the 16% in the new study. Its authors rated the certainty of the evidence across all studies as very low to moderate, and noted limited evidence among women.

What the new study adds to this is scale applied to the prediction question. The earlier large cohorts showed that low fitness goes with early death. This one tests, in more than half a million people, whether knowing fitness changes the estimate a clinician would otherwise make.

What the fitness and mortality study cannot show

Prediction, not cause. The study shows that a fitness measurement carries information about who will die. It does not show that a person who raises their fitness lowers their risk by the amounts given, and low fitness can be a sign of illness that has not yet been diagnosed, as well as a contributor to it.

People who had a treadmill test. Participants came from clinical exercise testing. People are usually given such tests for a medical reason, so the group may be older and less healthy than the general public, and an adjusted 10-year death risk of 13.4% even in the fit group points to a population at fairly high risk overall.

A low bar for unfit. The cut-off of 5.0 METs marked the bottom fifth of this particular group. In a younger or healthier population the least-fit fifth would sit higher, and the 32.9% figure would not carry over.

Six comparison factors. The standard model held age, sex, high blood pressure, diabetes, blood fats and smoking. Other things that affect survival and fitness alike, such as existing heart or lung disease, were not in the list of inputs, so some of fitness’s apparent contribution may stand in for them.

Estimated fitness. Fitness was taken from how far people got on a treadmill. Treadmill METs are generally estimated from the speed and slope reached, which is less exact than measuring oxygen use directly.

Death from any cause. The outcome was all deaths. The study does not say which causes of death fitness helped to predict.

One cohort. The results come from a single group of people, and risk models are usually checked in a second, separate group before they are relied on. The 2022 veterans study came from the same research group, so it is not confirmation by an independent team.

Whether an exercise test, or a fitness estimate, would change anything for a particular person is assessed by their doctor.

In half a million treadmill records, fitness added information beyond age, sex, blood pressure, diabetes, cholesterol and smoking, though less than the gap between the fit and unfit groups might suggest.