SYNTH-24 · Guide
The metabolic ceiling
Sustained energy expenditure tops out near 2.5 × BMR, and rest days are what bring a year back down to it
Primary sources
Thurber, Dugas, Ocobock et al., 2019 — *Extreme events reveal an alimentary limit on sustained maximal human energy expenditure.* Science Advances 5(6):eaaw0341. DOI 10.1126/sciadv.aaw0341 · REF-958, B
Best, Sadhir, Hyatt & Pontzer, 2025 — *Ultra-endurance athletes and the metabolic ceiling.* Current Biology 35(20):5106–5110.e2. DOI 10.1016/j.cub.2025.08.063 · PMID 41118725 · REF-959, B
Van Hooren, Rietjens & Plasqui, 2026 — *Longitudinal assessment of total daily energy expenditure in professional cyclists supports a maximal sustainable metabolic ceiling.* Current Biology 36(1):151–160.e2. DOI 10.1016/j.cub.2025.11.038 · PMID 41389793 · REF-960, B
Key Findings
The ceiling is duration-dependent and flattens near 2.5 × BMR. Thurber 2019 (REF-958, B, observational compilation; sample size not in the abstract) compiled total energy expenditure and BMR from endurance events lasting 0.5 to 250+ days and added new data from adults running ~250 km/week for 20 weeks in a transcontinental race: sustained metabolic scope falls curvilinearly with event duration and plateaus below 3 × BMR; incorporating overfeeding studies, the authors find evidence for an alimentary energy-supply limit of ~2.5 × BMR, beyond which the shortfall is paid from body stores.Best 2025 (REF-959, B) restates the same model as falling from ~10 × BMR at one day to an asymptote of ~2.5 × BMR at about 28 weeks.
The limit is alimentary, not muscular. The 2.5 × BMR figure comes from adding overfeeding data — the authors' reading is that energy supply through the gut, not output, sets the limit (Thurber 2019 abstract, REF-958: "evidence for an alimentary energy supply limit"; intake and absorption were not measured in the two later studies).The more precise "~2.36 × BMR intake independent of duration" figure formerly quoted here is not in REF-958 (full text; not checked against the abstract). Thurber also notes that transcontinental-race data suggest humans partially reduce expenditure during long events to extend endurance. This is why Grand Tour riders lose weight rather than eat their way even (inference).
Elite ultra-athletes approach it but do not beat it as a group. Best 2025 (REF-959, B, observational), n=14 highly trained to world-class ultra-endurance athletes (12 M / 2 F, 37.2 ± 7.0 y), doubly-labelled water during competitions of ~24 h to ~13 days and during lower- and higher-workload training weeks, with predictive equations applied to training records: maximum metabolic scope for competitions and periods up to 12 weeks approached but never exceeded the ceiling; 2.43 ± 0.17 × BMR over 30 weeks (TEE 4,085 ± 642 kcal/day) and 2.39 ± 0.17 over 52 weeks (4,020 ± 641 kcal/day).Four individuals exceeded 2.5 at 30 and 52 weeks (max 2.74); the authors conclude that exceeding the duration-dependent limit is rare. The "up to 7.08 × BMR for a ~24-h effort" figure formerly quoted here is not in the abstract and is not repeated.
A single day can run far above it. Doubly labelled water over a hot 100-mile race gave one world-class runner 16,104 kcal in 14 h 19 min (18.8 kcal/min) against 6,720 kcal eaten (REF-1194, C+, case study).With the paper's resting equation and his measured fat-free mass, that race day comes to roughly 11 × his daily resting expenditure (our arithmetic) — at or slightly above the ~10 × BMR that the curve puts at one day (REF-959, which found competitions approached but never exceeded the curve; both figures approximate, his resting rate predicted by equation, not measured).The ~2.5 × figure is the curve's long-run asymptote (~28 weeks), not a limit on a single day.
A full racing year in pros lands on the number. Van Hooren 2026 (REF-960, B, observational; expenditure estimated, not measured), n=8 male UCI-registered pros, one year of power files converted to expenditure with equations validated for professional cyclists: PAL 2.52 ± 0.11 across the whole year (not significantly different from 2.5), and 2.70 ± 0.14 counting only cycling training days (P < 0.001).
That gap is the finding that matters to a coach. Days without cycling training are what pull 2.70 down to 2.52. The authors' own reading (REF-960): the elevated expenditure on training days "highlights the importance of rest days in restoring energy balance and enabling athletes to maintain performance". Riding days sit ABOVE the ceiling; the year comes back to it (2.52, not below 2.5) because of the days that are not ridden.
What happens past the ceiling: the deficit is covered by body stores (Thurber 2019, REF-958, B). That this proceeds as fat, then muscle, alongside energy-sparing responses (immune suppression, menstrual disruption, elevated fatigue driving more rest) is the base's inference from the low-energy-availability literature, not a finding of these three papers.
Practical Application
Frame recovery weeks physiologically, not morally. A down week is not lost fitness or weak character; for an athlete living near pro-level volume it is how a whole season comes back to the limit (observed, not tested, as rest days in 8 male professionals — REF-960, B).Transfer, not tested: an amateur's yearly expenditure sits far below 2.5 × BMR, so for them this is an analogy, not the reason a down week is needed. The base's own practice, not from a study: athletes who resist rest may respond to this argument better than to "you need to recover".
Sanity-check annual ambitions. An athlete planning a year at pro-level training volume is planning to live at a boundary that eight professionals, fully supported, averaged right at (2.52 × BMR; REF-960, B). The base's own reading, not a finding: a plan with no rest days built in is unlikely to be sustained.
Weight loss in very long events can be read as an energy-supply problem rather than a discipline problem (inference from Thurber's model, REF-958 — expenditure above what can be absorbed is paid from body stores; body mass in stage races was not measured in these three papers). Fuelling strategy raises the floor; it does not remove the ceiling.
Never quote 2.5 × BMR at an individual as a target or a limit. It is a group mean from small samples. Use it to shape the conversation about volume and recovery, not to score anyone.
Limitations
Van Hooren's expenditures are ESTIMATED from power files via validated equations, not measured with doubly-labelled water. Best's are measured, but extrapolated from event and training weeks to 30- and 52-week windows through predictive equations rather than observed across them.
BMR itself is predicted, not measured, in the cycling study — an error in the denominator moves every ratio.
Individual variation is real and under-studied: four of Best's fourteen sat above the line. That cycling may permit higher values than ultrarunning (less muscle damage) is the base's inference, not a claim checked against the abstract.
Provenance note: this synthesis was prompted by a lay article (Escape Collective, Matt de Neef, 22 Jul 2026) that reported the three studies and interviewed two authors. Every number above was checked against the primary papers' abstracts on 2026-09-27; two figures that were not in the abstracts have been removed or labelled. The article also miscites Thurber 2019 as being in *Science* — it is *Science Advances*.
How this entry was checked and changed
Content correction 2026-09-27
"Sustained intake landed at ~2.36 × BMR independently of event duration" -> not in the Thurber 2019 abstract (which gives ~2.5 × BMR with plateau below 3×); labelled full text, unchecked.
"Short efforts go up to 7.08 × BMR for a ~24-h effort" -> not in the Best 2025 abstract; removed. Best's method (DLW in competitions and training weeks, predictive equations applied to training records; "up to 12 weeks approached but never exceeded") added from the abstract.
"Fat, then muscle, alongside energy-sparing responses" and "Best expects sport-specific ceilings" -> labelled inference, not findings of the three abstracts.
Thurber's transcontinental data (~250 km/week for 20 weeks; partial reduction of TEE during long events) and Van Hooren's "not significantly different from 2.5" added from the abstracts.
2026-10-01: cross-reference repointed from a merged framework to the one that replaced it (SYNTH-01→31, 05→34, 09→70, 12/15→32, 19→26).
Verification 2026-10-01: no refId-bearing claims exist in this framework — its three sources are not entries in the base.Checked directly against the PubMed abstracts: 2.43 ± 0.17 / 2.39 ± 0.17, TEE 4,085 ± 642 / 4,020 ± 641, four athletes > 2.5 (max 2.74), n=14 (12 M / 2 F, 37.2 ± 7.0 y), ~10× → ~2.5× at ~28 weeks, PAL 2.52 ± 0.11 and 2.70 ± 0.14 (p < 0.001), n=8 male UCI pros, BMR estimated, the rest-days quotation, Thurber's 0.5-250+ days / ~250 km/week / plateau below 3× / ~2.5× alimentary limit — all match.
Verification 2026-10-01: the numbers are real; the overreach is in the framing — 'independent', 'set by the gut' as fact, 'under' a ceiling the year average actually sits on, and a pro-level ceiling used to justify recovery weeks for everyone without a transfer clause.
Verification 2026-10-01: recommend filing the three papers as REF entries so the framework can cite refIds. Cross-reference (not audited per spec) describes SYNTH-34 as 'Taper for 4-6 h endurance / gravel events' while that entry's title is now 'The taper — cut the volume hard, keep every hard session...'.
Revision 2026-10-01: the three primary papers are now entries — Thurber 2019 REF-958, Best 2025 REF-959, Van Hooren 2026 REF-960, each level B, each written from the abstract. Every author-year claim in Key Findings, Practical Application and Limitations now carries its refId and level; the three entries were added to Cross-References and the Type line.
Revision 2026-10-01: numbers re-checked against the new entries' abstract-based text — all match (plateau below 3× and ~2.5× BMR; ~250 km/week for 20 weeks; ~10× → ~2.5× at ~28 weeks; n=14, 12 M / 2 F, 37.2 ± 7.0 y; 2.43 ± 0.17 / 2.39 ± 0.17; 4,085 ± 642 / 4,020 ± 641 kcal/day; four athletes above 2.5, max 2.74; n=8; PAL 2.52 ± 0.11 and 2.70 ± 0.14, P < 0.001). The ~2.36 × BMR figure is not in REF-958 and keeps its 'full text; not checked against the abstract' label.
Revision 2026-10-01: wording narrowed to what the entries carry — the rest-day link in REF-960 is observed, not tested ('measured as rest days' → 'observed, not tested'); REF-958's abstract gives no sample size, now said in Limitations. The stale description of SYNTH-34 in Cross-References was corrected. Level unchanged (B): three small observational studies.
Revision 2026-10-26: REF-1194 (Mougin 2026, C+, case study) added as the single-day end of the curve: 16,104 kcal in 14.3 h, roughly 11 × the daily resting expenditure by our arithmetic; the long-run ceiling and level unchanged.