REF-1560 · Lang et al., 2024 · British Journal of Sports Medicine · Paper · PubMed

Does fitness predict a longer life?

High vs low fitness: all-cause death HR 0.47 to 0.59, about 11-17% lower per 1 MET. Cohorts, usually one test, mostly men, no athletes: a strong predictor, not proof training buys it.

Main caveat: Cohort evidence only, with fitness usually measured once; change in fitness is not what most of it tests.
moderate evidenceMeasured on non-athletes and recreational people

The authors searched five databases to March 2024 and took each outcome from reviews with little overlap in their primary studies, choosing the best-rated review where overlap was high. They split evidence into mortality in general populations (8 reviews, 95 cohorts), new disease (9 reviews, 63 cohorts) and prognosis with existing disease (9 reviews, 51 cohorts).

Fitness predicted almost every outcome. The authors say the results may signal a causal relationship, and call fitness strongly protective; both are their reading. They also state that the evidence was limited to associations with a single fitness measure.

Protocol

Eligible reviews: at least two databases searched, pooling prospective or retrospective cohort or case-control studies of adults (18 years and over). Excluded groups: athletes, people recovering from surgery, disease at birth, pregnancy.
Exposure: measured VO2max, exercise tests without gas analysis, or non-exercise prediction equations.
Certainty: modified GRADE. Observational evidence started at "high" because trials were judged not feasible, an unusual starting point that flatters the ratings (our reading); it still ended very low to moderate.
Funding: no specific grant; competing interests none declared.

Key Results

Outcome (general populations unless stated)High vs low fitness, HR (95% CI)Per 1 MET higher, HR (95% CI)
All-cause mortality0.47 (0.39 to 0.56) to 0.59 (0.52 to 0.66)0.89 (0.86 to 0.92) to 0.83 (0.78 to 0.88)
Sudden cardiac mortality—0.49 (0.33 to 0.73); 2 studies
All cancer mortality—0.93 (0.91 to 0.96)
New heart failure0.31 (0.19 to 0.49)0.82 (0.79 to 0.84)
New hypertension0.63 (0.56 to 0.70)—
New stroke—0.97 (0.96 to 0.98)
Cardiovascular death, people with cardiovascular disease0.27 (0.16 to 0.48)—
Adverse events, pulmonary hypertension0.81 (0.78 to 0.85)—
Dashes in the table: the estimate is in the paper's figures, not its text, or was not reported; a dash is not a null. Sex: 1 858 274 men against 180 202 women for all-cause mortality. Authors report similar sex patterns, with wider CIs for women.
Prostate cancer: HR 1.15 (1.00 to 1.30) for high fitness in men, which the paper calls null.
Certainty: very low to moderate for mortality, very low to low for new disease and for prognosis in clinical groups. Heterogeneity was often high and unexplained.
Reporting: the sentence on heart rate recovery says delayed recovery went with an 83% reduced risk of adverse events, which reads as the wrong way round (our reading). Group counts of 95, 63 and 51 cohorts sum to 209 against 199 unique (our arithmetic), which overlap between groups could explain.

Mechanisms

Measured: nothing mechanistic.
Authors' description: fitness reflects the integrated function of many body systems.
Inference: that could be why it predicts so widely, but the same breadth fits confounding. Illness, genes, smoking and weight all lower fitness and raise risk, so a broad gradient cannot by itself separate cause from marker.

Practical Application

Does fitness predict a longer life? High vs low fitness: all-cause death HR 0.47 to 0.59, about 11-17% lower per 1 MET. Cohorts, usually one test, mostly men, no athletes: a strong predictor, not proof training buys it.
Good for: explaining why fitness is worth keeping for someone training for health with no race, and why a move from unfit to fit is where the authors locate the largest gains.
Not good for: telling a trained rider that another 1 MET buys the same percentage. The paper prints no curve shape and only cites the largest gains for a move from unfit to fit, and athletes were excluded.
Not good for: a percentage of life gained. These are relative hazards at group level, and no trial is behind them; see REF-1562 and REF-1563.

Limitations

Cohort evidence only, with fitness usually measured once; change in fitness is not what most of it tests.
Review quality: 11 critically low, 4 low, 8 moderate, 3 high on AMSTAR2.
Reverse causation and confounding: undiagnosed illness lowers fitness, and adjustment was uneven, especially in clinical groups.
Mixed exposure: direct, estimated and non-exercise fitness are pooled across reviews. Women are sparse, and athletes absent.
How this entry was checked and changed
Why this level

an overview of meta-analyses of cohorts is not a meta-analysis of controlled trials, so its size does not lift it above the observational cap; 11 of the 26 reviews were rated critically low quality on AMSTAR2 and only 3 high