The 2018 JAMA VO2 max study, dissected: what 122,000 patients actually showed

Almost every longevity article you’ve read that claims “fitness is the strongest predictor of how long you’ll live” is, whether it says so or not, leaning on one paper. It’s a 2018 study in JAMA Network Open, it followed 122,007 people, and it produced a number so clean that it now gets quoted like scripture: being in the fittest group was linked to an 80% lower risk of dying than being in the least-fit group.

It’s a genuinely important study. It’s also the most misquoted one in the whole longevity space, because the 80% gets ripped out of context and turned into a personal promise it was never designed to make. So let’s do the thing nobody bothers to do with the studies they cite: actually read it. Design, numbers, and — the part the headlines skip — where it stops.

If you want the plain-English version of why VO2 max matters at all, that’s a separate post: why cardiorespiratory fitness is the strongest predictor of longevity. This one is about the evidence underneath that claim, and how much weight it can actually bear.

Why this study matters

Longevity science has a supply problem: most of the strong causal evidence comes from mice, and most of the human evidence is observational. In that landscape, a study earns its authority through sheer scale and a hard endpoint — and this one has both. 122,007 patients. Death as the outcome, not a surrogate marker. Over a million person-years of follow-up. When something is measured that many times against the least ambiguous outcome there is, the signal is worth taking seriously even before you argue about what causes it.

It also produced a soundbite that reframed how a lot of clinicians think: the authors reported that the mortality risk of poor fitness was comparable to or greater than the risk from coronary artery disease, smoking, or diabetes. That’s the line that launched a thousand VO2-max threads. Whether it means what people think it means is exactly what we’re here to check.

The design: who, how many, and what they actually measured

Kyle Mandsager and colleagues at the Cleveland Clinic pulled the records of 122,007 consecutive patients who underwent an exercise treadmill test between 1991 and 2014. Mean age was 53.4, and 59.2% were male. They linked each patient to mortality records and followed them for a median of 8.4 years, over which 13,637 people (11.2%) died.

Here’s the first detail that matters, and it’s the one nine out of ten summaries get wrong. They did not measure VO2 max. They measured peak estimated metabolic equivalents — METs — from the treadmill test. A MET is a multiple of resting energy use, and the treadmill software estimates it from the speed and gradient you reach before you stop. Mean peak was 9.0 METs; the elite group averaged 13.8.

Estimated METs and lab-measured VO2 max are cousins, not twins. Real VO2 max comes from a mask measuring the oxygen in your breath while you go to exhaustion. Treadmill-estimated METs assume you’re as economical as the average person at a given speed and grade, which isn’t always true — the estimate can be off by 10-15% for an individual. So when you see this study cited as “the VO2 max study,” treat that as a fair shorthand for cardiorespiratory fitness, not a claim that anyone wore a gas-exchange mask. It matters because it caps how precisely the result applies to your watch’s VO2 max estimate, which is itself an estimate of an estimate.

The clever bit of the design was the categories. Rather than one fitness threshold for everyone, they ranked people within their own age and sex and split them into five percentile bands:

• Low — below the 25th percentile
• Below average — 25th to 49th
• Above average — 50th to 74th
• High — 75th to 97.6th
• Elite — 97.7th percentile and up

That age-and-sex-relative framing is why a fit 65-year-old woman and a fit 30-year-old man can both land in “elite” despite very different raw numbers. It’s the right way to do it, and it’s why the results are about relative fitness, not an absolute MET target.

What they found

The headline holds up. After adjusting for the usual suspects — age, sex, coronary artery disease, diabetes, hypertension, smoking, medications, and more — the fittest group had a dramatically lower death rate.

• Elite vs low fitness: hazard ratio 0.20 (95% CI 0.16–0.24). That’s the ~80% lower risk everyone quotes.
• Below-average vs above-average: hazard ratio 1.41 — so even a middle-of-the-pack drop in fitness came with a 41% higher risk.
• Elite vs high: hazard ratio 0.77 (p = 0.02).

That last one is the quietly radical finding. Going from “high” fitness (already top quartile) to “elite” still cut risk by nearly a quarter. This is what the authors meant by no observed upper limit of benefit — unlike, say, blood pressure or LDL, where extremely low can stop helping or start hurting, fitness kept paying out at the very top of the distribution. There was no point in this dataset where being fitter stopped being associated with living longer.

And then the comparison that made the study famous. When they lined up low fitness against the classic risk factors, being unfit carried a relative risk on par with or worse than end-stage kidney disease, smoking, coronary artery disease, and diabetes. Sitting still, in other words, sat comfortably among the heavy hitters of preventable mortality.

If you want to see how this stacks against every other predictor people obsess over, I put them side by side in the longevity biomarkers ranked by how well they predict death — fitness comes out at or near the top, and this is the study doing most of the lifting.

The caveats the headlines skipped

Right, so this is the part that gets cut for length in every viral version. A big, clean result deserves a big, honest asterisk — actually four of them.

1. It’s observational, so reverse causation is the ghost in the room. The study shows fitness and survival move together. It cannot, by design, prove that raising your fitness lowers your risk. The most important alternative explanation is that early, undiagnosed illness lowers your fitness and raises your death risk at the same time. A person with a quietly failing heart or a not-yet-found cancer will score low on a treadmill and also die sooner — not because low fitness killed them, but because the same underlying disease did both. In a population referred for cardiac testing, that’s not a hypothetical; it’s a live concern. Statistical adjustment helps but can’t fully remove it, because you can’t adjust for a disease nobody has diagnosed yet.

2. It was a cardiac-referral population, not the public. These were people sent for a stress test — often because something prompted the referral. Mean age 53, most being worked up for symptoms or risk. That’s a specific slice of humanity, and the exact hazard ratios don’t automatically transfer to a healthy 28-year-old marathoner. The direction of the effect is one of the most reproduced findings in exercise science; the precise 0.20 is a property of this cohort.

3. It’s a single snapshot. Fitness was measured once. The study can’t see who was on the way up and who was on the way down, or whether ten years of training beats a lucky test on a good day. A one-off treadmill result is a decent proxy for someone’s typical fitness, but it isn’t a trajectory.

4. Unmeasured confounders. The authors are upfront that they lacked data on socioeconomic status, race and ethnicity, diet, and other lifestyle factors. Fit people tend to differ from unfit people in dozens of ways that also affect longevity, and no adjustment model catches all of them.

None of this makes the study wrong. It makes it observational — which is exactly as strong as observational evidence gets, and no stronger.

Where the evidence gets stronger — the cross-study read

Here’s the synthesis you won’t get from reading this paper alone, and it’s the thing that actually moves how much to trust it.

The reverse-causation objection — “maybe sick people are just unfit” — sounds fatal until you ask a different question: what happens to people who change their fitness? If low fitness were only a passive marker of hidden illness, then improving your fitness shouldn’t move your risk. But it does. Steven Blair’s Aerobics Center Longitudinal Study (JAMA, 1995) followed men across two fitness tests years apart and found that those who went from unfit to fit had substantially lower mortality than those who stayed unfit. Erikssen and colleagues (The Lancet, 1998) found the same thing over seven years in Norwegian men. Change the input, change the outcome — that’s harder for a pure reverse-causation story to explain than a single snapshot is.

But be honest about what those studies are: they’re observational too, not randomised trials. People whose fitness improved could differ from those who stayed unfit in ways that also affect mortality — baseline health, a new diagnosis, healthcare access, other habits that travel with taking up exercise. So the change studies weaken the reverse-causation objection; they don’t kill it, and they don’t prove the fitness change itself is what lowered the risk. The only thing that would settle causation cleanly is a randomised trial with a mortality endpoint, and for lifelong fitness that trial is effectively impossible to run.

And the magnitude generalises. Kodama and colleagues pooled 33 studies in a JAMA meta-analysis (2009) and found each 1-MET increase in cardiorespiratory fitness was associated with roughly a 13% lower all-cause mortality risk. That’s the number I’d actually anchor on, because it’s an average across many populations rather than one referral cohort — and it lines up neatly with the Cleveland data once you convert the MET gaps into risk. (It’s also, still, an association — a bigger and more consistent one, but the same species of evidence.)

So the honest position isn’t “the 2018 study proves fitness saves your life” and it isn’t “it’s just correlation, ignore it.” It’s this: one enormous observational study, plus change-over-time studies that weaken the reverse-causation objection, plus a multi-study meta-analysis that reproduces the dose-response — together those make cardiorespiratory fitness about as well-supported a modifiable longevity lever as we have in humans, which is a real claim precisely because it stops short of proven causation. No single paper gets you there. The stack does, and it’s still a stack of associations — strong enough to act on, not strong enough to promise.

What it actually means for you

The thesis, stated plainly: this study is the strongest single piece of evidence that fitness tracks survival — but it measured one treadmill test in cardiac-referral patients, so it proves fitness is a powerful risk marker, not that pushing your personal number to “elite” hands you back the full 80%.

Which leads to a genuinely useful decision framework:

• Treat a low fitness score as a warning light, not a footnote. This is the study’s most defensible, most actionable message. If you’re in the bottom quartile for your age and sex, that’s information on par with a bad cholesterol panel — worth acting on now.
• Chase the first few METs hardest. Because there’s no upper limit doesn’t mean every zone is equal value. Going from low to above-average is where the risk curve is steepest and the return on effort is largest. The couch-to-competent jump matters more than the good-to-elite one.
• Don’t over-read the 80% as a personal promise. It’s a between-group, population-level, relative number in a specific cohort. Build fitness because the wider evidence — including the change studies — says improving it genuinely helps, not because one hazard ratio guarantees you two extra decades.
• Build it the boring way. The mechanism doesn’t care how you got fit. A Zone 2 base with a weekly hard session is the most reliable route, and I’ve argued the Zone 2 versus intervals split elsewhere — the short version is you need both, and most people underdo the easy end.

Genuinely, that’s the whole thing. A landmark study, quoted half-right by nearly everyone, that says something slightly humbler but far more useful than the soundbite: your fitness is one of the loudest signals you have about where your health is heading, and unlike most such signals, you get a vote in it.

Bit nerdy, but that’s the point of reading the actual paper instead of the tweet about it.