REF-1561 · Kokkinos et al., 2022 · Journal of the American College of Cardiology · Paper · PubMed

Is very high fitness risky?

In 750,302 US veterans, death risk was lowest near 14 METs (HR 0.24, comparator unnamed in the abstract), with no rise above it. Veterans, cohort, not proof, and death is the only outcome reported here.

Main caveat: Abstract only. Covariates, exclusions, reason for testing and early-death handling were not read.
moderate evidenceMeasured on non-athletes to trained athletes

Categories (quintiles and the 98th percentile) were set within age and sex from peak METs on the test. There were 174,807 deaths over 7,803,861 person-years, or 22.4 per 1,000 person-years. The association was inverse and graded across age, including 110,637 septuagenarians and 26,989 octogenarians, and across race and sex.

The authors conclude that being unfit carried a greater risk than any cardiac risk factor they examined. That is a statement about prediction; the abstract gives no figures for it.

Protocol

Who: US veterans aged 30 to 95, including 142,798 African American, 35,197 Hispanic and 16,050 Native American participants and 45,232 women, about 6.0% of the sample (our arithmetic).
Exposure: peak METs on a standardised treadmill test. The abstract describes no repeat tests and does not say whether METs were measured by gas exchange or estimated from workload.
Analysis: multivariable Cox models; covariates are not in the abstract.
Not read: the paper itself, its exclusions, handling of early deaths and funding.

Key Results

ComparisonHR (95% CI)
About 14.0 METs, men0.24 (0.23 to 0.25)
About 14.0 METs, women0.23 (0.17 to 0.29)
Least fit (20th percentile) vs extremely fit4.09 (3.90 to 4.20)
Comparator: the abstract does not state it for the 14-MET figures. 1/4.09 is 0.24 (our arithmetic), which suggests the least-fit group, but this is our reading.
Scale: at the usual 3.5 mL/kg/min per MET, 14 METs is about 49 mL/kg/min (our arithmetic), if the MET value is taken at face value.
Reporting: 4.09 is not central in its CI on a ratio scale (3.90 is 1.05 times below, 4.20 is 1.03 times above; our arithmetic), so the printed interval may contain a slip.

Mechanisms

Measured: nothing mechanistic in the abstract.
Inference: a person ill enough to stop early on a treadmill scores a low MET value and is also more likely to die, so part of this gradient may run from illness to fitness. Very high fitness also travels with favourable genes and few diseases.

Practical Application

Is very high fitness risky? In 750,302 US veterans, death risk was lowest near 14 METs (HR 0.24, comparator unnamed in the abstract), with no rise above it. Veterans, cohort, not proof, and death is the only outcome reported here.
Good for: answering whether extreme fitness showed excess mortality in a huge sample: it did not. Absence of a rise is the finding; the abstract reports no formal test of equivalence.
Good for: the low end. The least fit carried the highest risk; an athlete is usually far up this scale (inference).
Not good for: heart rhythm or other non-fatal harms of long endurance careers; see REF-1577.
Not good for: a target of 14 METs. It is where this curve bottoms out, not a tested goal.

Limitations

Abstract only. Covariates, exclusions, reason for testing and early-death handling were not read.
Baseline fitness only, as far as the abstract shows, so later change is invisible.
Selected population: veterans, few women; reverse causation and healthy-survivor effects are not excluded.
Associations: the authors' word "independent" means after statistical adjustment; it is not evidence of cause.
How this entry was checked and changed
Why this level

(from the abstract) — large observational cohort, at the observational cap however large it is; adjustment set, exclusions and the reason for testing could not be read