Transcript
Cardiorespiratory fitness is one of the most consistent predictors of how long a person lives. Across dozens of studies and hundreds of thousands of participants, the pattern holds: the fitter you are, the less likely you are to die in any given follow-up window. That is not a contested finding. What is contested — and worth thinking through carefully — is what that pattern actually means, whether exercise causes longer life or merely correlates with it, and what that uncertainty implies for the growing industry of longevity supplements that measure themselves against exercise as their gold standard.
Start with the numbers, because they are striking enough to deserve full attention before we complicate them.
A meta-analysis by Laukkanen and colleagues, drawing on 37 studies and roughly 2.25 million participants, found a pooled relative risk of 0.55 when comparing the top third of cardiorespiratory fitness against the bottom third for all-cause mortality. A relative risk of 0.55 means the high-fitness group was roughly half as likely to die during the study's follow-up period as the low-fitness group. Half as likely. Across 2.25 million people. That is a large and consistent signal by any standard in epidemiology.
A separate line of evidence from the Kokkinos study — which tracked more than 750,000 United States veterans — found that each one-MET increase in fitness was associated with a 13 to 15 percent reduction in mortality risk. A MET is a unit of metabolic output. And a scoping review reported by Physiology Friday, covering more than 20.9 million observations, found that high cardiorespiratory fitness compared with low was associated with a 41 to 53 percent lower risk of premature mortality.
One number from that veterans cohort deserves particular attention. The least-fit individuals carried a hazard ratio of 4.09 compared with the most fit — meaning they were roughly four times as likely to die during follow-up. To put that in context: the study's authors concluded that being unfit carried a greater risk than any of the cardiac risk factors they examined. Being in the lowest fitness group was associated with a far greater mortality risk than either of those two conditions, which public health has spent decades trying to address. And it is a legitimate benchmark. But understanding what kind of benchmark it is requires sitting with some uncomfortable caveats.
Because here is what Laukkanen's meta-analysis also reports, and this is the part that rarely makes it onto wellness podcasts: the GRADE quality rating for this entire body of evidence is "very low." Not low. Very low. The researchers flagged serious risk of bias, substantial heterogeneity across studies — an I-squared above 90 percent, meaning the studies varied enormously in their results — and concerns about publication bias. The relative risk of 0.55 is a real number from a real meta-analysis of 2.25 million people. It is also, by the standards of evidence grading that clinical medicine uses, not as solid as it sounds.
Why the discrepancy between scale and quality? Because almost all of these studies are observational. They measure who is fit and who dies, but they cannot separate the effect of fitness itself from the many other ways in which fit people differ from unfit people. People who have high cardiorespiratory fitness are often people who have never had serious illness, who have favourable metabolic profiles, who make different dietary and lifestyle choices, who have better access to environments that support movement. Fitness may be less a cause of longevity and more a marker of the kind of overall biological health that also produces longevity.
Think of it this way. People who carry umbrellas are rarely caught in the rain. Umbrellas and dry clothes correlate strongly. But cautious, organised, weather-aware people are also the ones who check forecasts, dress sensibly, and plan their routes. The umbrella is real. The correlation is real. But the umbrella alone is not the whole story.
A 2024 study published in The Journal of Clinical Endocrinology and Metabolism, described by the primary-md.com blog, pushed this logic further using a technique called Mendelian randomisation. The idea is that certain genetic variants influence a trait — in this case, VO2max — in a way that is essentially randomly assigned at conception, unaffected by later lifestyle choices. If higher VO2max actually causes longer life, then people who are genetically predisposed to higher VO2max should tend to live longer. That is the prediction. The finding was that genetically predicted VO2max showed no association with longevity. None. That is a striking result, and it suggests that at least some of the observational association reflects confounding — that people who happen to have high aerobic fitness were already healthier for other reasons, genetic and environmental, before they ever started exercising.
There is also a structural problem with many of the largest observational studies. The Cleveland Clinic cohort, published in JAMA Network Open in 2018, which found that elite performers had about 80 percent lower mortality than the lowest-fitness group, drew its participants from people referred for clinical stress testing. Many in the low-fitness group were there precisely because of suspected cardiovascular problems. That selection makes the low-fitness group sicker from the outset, which likely inflates the apparent mortality gap between fit and unfit.
So hold two things simultaneously. The observational evidence that fitness predicts longevity is enormous in scale and consistent in direction. The relative risk of 0.55, the hazard ratio of 4.09, the 41 to 53 percent lower premature mortality risk — these numbers come from serious research involving millions of people. And the GRADE rating for this evidence is very low quality. And the 2024 Mendelian randomisation study found no causal signal. Both things are true. The number is real; the causal interpretation is genuinely uncertain.
This is why randomised controlled trials matter. In an observational study, you cannot stop some people from being healthy and force others to be unfit. You can only watch what happens. A randomised trial takes people and assigns them, at random, to exercise more or to continue as they are. If the exercise group lives longer, that is much stronger evidence of a causal effect, because randomisation should balance out the pre-existing differences between groups.
The problem is that running a randomised trial of exercise for mortality is extraordinarily difficult. You cannot blind people to whether they are exercising. You cannot prevent control participants from independently deciding to move more. You need very large numbers and very long follow-up to detect mortality differences. And ethically, you cannot tell a control group to become deliberately sedentary.
Which brings us to Generation 100 — a large and long randomised controlled trial of exercise in a general population of older adults aged 70 to 77. Conducted in Norway, it randomised adults aged 70 to 77 into groups doing high-intensity interval training, moderate-intensity continuous training, or following standard Norwegian physical activity guidelines as controls. The trial ran for five years. The outcome measure was all-cause mortality.
The HIIT group did not significantly outlive the controls. The absolute difference in mortality was 1.7 percentage points, and it was not statistically significant. In a trial of this scale and duration, a difference of that size that does not reach statistical significance means we cannot rule out that it was due to chance. The trial's null result, published in the BMJ, is not a minor footnote. It is the strongest direct evidence we have on whether adding structured exercise to the lives of older adults measurably extends life.
There is an important wrinkle, though, and it matters for interpreting the result. Control participants in Generation 100 ended up being considerably more active than the trial had anticipated — they were following general guidelines and, it turned out, doing so reasonably well. The comparison was not between vigorous exercisers and sedentary people. It was between supervised exercisers and a control group that was already largely active, nearly a quarter of whom took up high-intensity training on their own. That may explain why no mortality difference emerged. Or it may mean that above a certain baseline of activity, adding more structured training does not move the mortality needle in older adults over a five-year window. The trial cannot distinguish between those two interpretations.
What Generation 100 does not tell us is that exercise is useless. It tells us something more specific and more interesting: that in adults who are already doing some activity, adding a formal high-intensity exercise programme on top of that did not produce a detectable survival advantage within five years. The intervention did produce real benefits — the HIIT group showed larger improvements in VO2max and quality-of-life measures than the moderate-intensity group, even if neither group's survival advantage over controls was statistically clear.
Now consider what this means for the supplement question — and for the specific framing of Zone 2 training.
Zone 2 refers to low-to-moderate intensity aerobic training, typically at an effort level where you can hold a conversation, usually described as 150 to 200 minutes per week. It has become a central piece of longevity discourse, championed by practitioners who recommend combining it with one or two high-intensity sessions weekly. The physiological rationale is real — this intensity trains mitochondrial efficiency and fat oxidation. But the claim that Zone 2 specifically, at this dose, reduces mortality is not supported by any randomised trial that isolates Zone 2 against other intensities for a mortality outcome. Generation 100, which is the closest thing we have to such a trial, actually found that HIIT outperformed moderate-intensity continuous training on VO2max and quality-of-life measures, even though neither beat controls on survival. The prominence of Zone 2 in longevity discussions reflects practitioner consensus and plausible mechanisms — not a randomised evidence base that isolates it from other training approaches. That is not a reason to dismiss it; moderate aerobic activity is supported by a broad base of evidence. The point is that the specific zone framing is, at least partly, practitioner shorthand that has been amplified into something that sounds more precisely evidence-based than it is.
The observational case that fitness predicts longevity is consistent, massive in scale, and associated with effect sizes that match or exceed classic risk factors such as smoking and diabetes. The hazard ratio of 4.09 for the least fit, larger than any traditional cardiac risk factor in the same analysis, is a genuine signal worth taking seriously. At the same time, the GRADE evidence quality is very low, the Mendelian randomisation evidence cuts against a simple causal story, and the only large randomised mortality trial in older adults found no statistically significant survival benefit from structured exercise. That is the honest state of the evidence.
Against that backdrop, consider NMN, resveratrol, spermidine, and the various other molecules marketed for longevity. Their proponents often use exercise as the baseline — the thing we know works — and argue their pill offers comparable or additive benefits. But if exercise itself has never demonstrated a mortality benefit in a randomised trial, the benchmark is not a cleared bar. It is a plausible, biologically credible intervention with very strong observational support and genuine mechanistic rationale that has not yet survived the rigour of a long-term randomised mortality trial.
None of the longevity supplements have evidence approaching even that uncertain standard. They typically have short-term trials in small numbers of people, often testing biomarker changes rather than mortality outcomes. Resveratrol's early promise in animal models did not translate to consistent effects in human trials. NMN has human safety data and some metabolic signals, but no mortality data. Spermidine has observational associations and interesting cellular mechanisms around autophagy, but again, no randomised mortality trial. The gap between supplement evidence and exercise evidence is not small. Even with all the caveats about observational data and Mendelian randomisation, the cardiorespiratory fitness literature now spans 37 cohort studies involving more than two million participants.
The question of what evidence bar a longevity pill should clear before being taken seriously has a logical structure, even if it does not have a simple answer. If the claim is that the pill extends life, the evidence should include at minimum a well-powered randomised trial with mortality or validated surrogate outcomes, in humans, with appropriate duration. That standard has not been met by any longevity supplement currently on the market. The observational evidence for exercise, despite its limitations, dwarfs anything available for any supplement in this space.
The genuinely open question is whether the mortality benefit from fitness in observational studies reflects exercise behaviour itself — the sheer volume of habitual physical activity over a lifetime — rather than aerobic capacity as a physiological variable. The Mendelian randomisation finding suggests that VO2max as a trait may be less causally important than what people do with their bodies day to day. If that is right, it shifts the emphasis from optimising fitness metrics to sustaining habitual movement — a distinction that changes what training for health actually means. It also leaves open whether Generation 100's null result applies only to older adults who are already active, or whether people starting from lower baselines would show a different answer. Those questions are not resolved.
What the evidence does establish, across its full range of quality and uncertainty, is that being consistently physically active across a lifetime is the most robustly associated behavioural predictor of longer life we have. The causal mechanism is uncertain enough that the specific numbers should be held carefully rather than cited as proof. And the supplement evidence, however it is framed, has not come close to meeting even that uncertain standard. Any pill that wants to compete seriously with exercise for a place in a longevity strategy needs randomised mortality data in humans. None of them have it yet.