REF-959 · Best et al., 2025 · Current Biology, Vol. · Paper · PubMed

Ultra-endurance athletes and the metabolic ceiling

Blimited evidenceMeasured on highly trained to world-class athletesWritten before October 2026, being re-checked

Thurber and colleagues proposed that the human ceiling on sustained energy expenditure depends on duration: it drops semi-logarithmically from ∼10× BMR for 1-day events, levelling off near ∼2.5× BMR by approximately 28 weeks. Here the model was tested in 14 ultra-endurance athletes ranging from highly trained to elite and world-class.

Doubly labelled water gave metabolic scope (MS; total energy expenditure / BMR) during ultra-endurance races lasting ∼24 h to ∼13 days and during lighter and heavier training weeks; from these the authors built equations predicting MS from exercise volume and ran them over each athlete's training log to find peak MS across windows of 1, 3, 6, 12, 30 and 52 weeks.

Up to 12 weeks, peak MS in races and training came close to the proposed ceiling but never went past it. At 30 and 52 weeks four athletes went above 2.5× BMR (maximum: 2.74), but the sample mean did not.

Participants

n=14 highly trained, elite and world-class ultra-endurance athletes (12 males and 2 females; 37.2 ± 7.0 years).

Protocol

Metabolic scope (TEE/BMR) came from doubly labelled water, measured in ultra-endurance races lasting ∼24 h to ∼13 days and in lighter and heavier training weeks.
From these data the authors derived equations that predict MS from exercise volume.
Running the equations over each athlete's training log gave peak MS for windows of 1, 3, 6, 12, 30 and 52 weeks.
Comparison: the duration-dependent ceiling proposed by Thurber and colleagues (∼10× BMR at 1 day, asymptote of ∼2.5× BMR at approximately 28 weeks).

Key Results

PeriodMetabolic scope (× BMR)TEE
Competitions and training periods up to 12 weeks (maximum MS)approached, but never exceeded, the proposed ceiling—
30 weeks (sample mean)2.43 ± 0.174,085 ± 642 kcal/day
52 weeks (sample mean)2.39 ± 0.174,020 ± 641 kcal/day
At 30 and 52 weeks, four athletes went above 2.5× BMR (maximum: 2.74), yet the group means at 30 and 52 weeks stayed within the proposed limit.
The authors' reading: exceeding the hypothesized duration-dependent limit of energy expenditure is rare.

Practical Application

In 14 highly trained to world-class ultra-endurance athletes (doubly labelled water in competitions and training weeks, then equations applied to training records), the highest metabolic scope for competitions and for periods up to 12 weeks approached but never exceeded the proposed duration-dependent ceiling, and the group averaged 2.43 ± 0.17 × BMR over 30 weeks and 2.39 ± 0.17 over 52 weeks — below the ∼2.5× BMR limit, although four athletes were above it (maximum 2.74). To "can a very fit athlete train past the ceiling for a whole year?": as a group these athletes did not; a few individuals sat slightly above it.
The individual spread matters: four of fourteen were above 2.5× BMR over 30 and 52 weeks, so the figure is not a hard personal limit.
Inference, not tested here: athletes training far less than these sit well below the ceiling, so for most amateurs it is not the binding constraint on their year.
Not from this paper: the model itself is REF-958; a year of professional cycling against the same ceiling is REF-960; the base's use of both is SYNTH-24.

Limitations

n=14, only two women; ultra-endurance athletes of mixed level, sports not listed in the abstract.
The 1- to 52-week values are calculated by applying predictive equations to training records; expenditure was measured only during the competitions and selected training weeks.
Whether BMR was measured or predicted is not stated in the abstract.
No energy intake, body mass or health outcome is reported in the abstract, so it cannot say what happened to the four athletes above the limit.
The model's originator is a co-author of this test of it.
Written from the abstract and record metadata only (full text not read).
How this entry was checked and changed
Why this level

small cohort, n=14; the 30- and 52-week values are calculated from training records, not measured across those periods