Anaerobic and aerobic energy system contribution during maximal exercise: a systematic review
102 studies; the 50/50 point is 78.6 s, beyond ~80 s all-out efforts are majority-aerobic (updates Gastin 2001)
Background
Gastin's 2001 review (our REF-185) established the classic figures for relative aerobic/anaerobic contribution by effort duration. This systematic review consolidates evidence from multiple methodologies to give a more precise picture of how the phosphagen, glycolytic and oxidative systems share ATP resynthesis during single bouts of maximal exercise of different durations.
Methods (abstract)
Systematic search of seven databases (CINAHL, Cochrane, Embase, MEDLINE/PubMed, Scopus, SPORTDiscus, Web of Science), studies from 1984 to January 2020
Peer-reviewed English-language studies of energy-system contributions during maximal exercise in adults (≥18 years)
Studies categorized by assessment approach: oxygen deficit (OD, 66 studies), mixed methods (MM, 33), theoretical models (TM, 7)
Nonlinear regression to estimate anaerobic and aerobic contributions across exercise durations
Key Findings (abstract)
n=102 studies, 311 data points (study means from samples that were 78% male, 12% female, 11% mixed)
The anaerobic system predominates in short maximal exercise up to approximately 75-80 s
Equal anaerobic and aerobic contribution at 78.6 s (95% CI ± 1.1 s); longer durations draw an increasingly greater proportion of aerobic energy
No differences between running and cycling, or by training status, in the regression curves
Small but significant effects of measurement method and pacing strategy
Per-duration percentage shares (the aerobic fraction at each specific effort length) are not reported in the abstract; the earlier table of such values was removed
Practical Application
Systematic review of 102 studies (311 data points, 78% male samples, literature 1984–Jan 2020): in a single all-out bout the anaerobic system predominates only up to about 75-80 s — the point of equal contribution was 78.6 s (95% CI ± 1.1 s) — and beyond that the aerobic share grows with duration; the curves did not differ between running and cycling or by training status, with only small effects of measurement method and pacing. Meaning: any maximal effort longer than about 80 s is majority-aerobic, which is why aerobic capacity matters even for events athletes experience as "anaerobic".
Event-specific periodization: efforts under ~1 min are anaerobic-dominant; efforts of several minutes and beyond are aerobic-dominant — so pursuit-, middle-distance- and criterium-type events need both systems, with the aerobic side carrying most of the energy (the exact split per duration is not in this abstract)
Misconception corrections supported by the abstract: "sprint events are all anaerobic" is false for anything approaching the 78.6 s crossover; "VO₂max doesn't matter for short events" is false once the effort passes ~80 s
Authors' conclusion: tailor high-intensity training, work-rest ratios and pacing to improve anaerobic capacity, oxygen-uptake kinetics and short-term recovery
Because training status did not shift the curves, the same duration logic applies to amateurs and elites
Updates vs Gastin 2001
Larger evidence base (102 studies) and a formal nonlinear regression across durations
Measurement method (OD vs mixed vs theoretical) and pacing have small significant effects — a reason older single-method figures differ between sources
Sex balance remains poor (12% female samples)
Whether the 2001 estimates were "confirmed within X%" is not stated in the abstract (earlier claim removed)
Limitations
Systematic review, not new primary data; underlying studies are measurement studies, not controlled trials
Measurement-method heterogeneity (OD / mixed / theoretical) has a significant effect on estimates
78% of samples male; 12% female
Heavy reliance on lab-based maximal tests (less real-world race data)
Pacing strategy affects contributions — all-out vs evenly-paced bouts are not interchangeable
How this entry was checked and changed
Why this level
regraded 2026-09-27 from A: systematic review of measurement studies (not controlled trials) with acknowledged method heterogeneity
Content correction 2026-09-27
Duration-by-percentage table (10 s 10-15% … >60 min ~99%), "4-min efforts ~80% aerobic", "crossover ~50-60 s", "TT 95-97%", "confirmed within ~5%" -> not in the abstract; removed. Abstract: anaerobic predominance up to ~75-80 s, equal contribution at 78.6 s (95% CI ± 1.1 s).
Added the abstract's methods and counts (102 studies, 311 data points, 78/12/11% male/female/mixed, OD 66 / MM 33 / TM 7, 1984–Jan 2020) and the null result for running vs cycling and training status.
Level A -> A- (systematic review of measurement studies with method heterogeneity); added Tier/Sport lines. Cross-ref to REF-186 reworded (that entry documents a pyramidal TID, not an "80% aerobic" figure).
2026-10-01: cross-reference to SYNTH-13 no longer names an archetype — that model was retired as an unvalidated composite.
Full text (checked 2026-10-12)
Modelled aerobic share of a single maximal bout: 0–10 s 9%, 0–30 s 25%, 0–60 s 42%, 0–75 s 49%, 0–90 s 54%, 0–120 s 62% (full text, Table 5).
For longer maximal efforts the modelled aerobic share was 72% at 0–180 s, 78% at 0–240 s, 81% at 0–300 s, 89% at 0–600 s and 94% at 0–900 s (full text, Table 5).
The estimates differ from Gastin 2001 by 0 to 3% for any given duration (2–3% more aerobic up to 15 s, 1–3% more anaerobic for 30 to 240 s), and 28 studies (27%) were carried over from the original review (full text).
The regression model has a prediction error of about 12–14%, which the authors say requires cautious interpretation of the crossover point and of contributions at specific durations (full text).
Curves did not differ between running and cycling (p = 0.793) or training status (p = 0.355); the oxygen-deficit curve differed from mixed methods and theoretical models (p = 0.001), and constant-intensity exercise differed from all-out and self-paced efforts (p = 0.012) (full text).
Samples ranged from 1 to 32 participants (most often 8), aged 18–41 years (mean 24.4 ± 4.9 years), and were described as trained (78%), untrained (20%) or mixed (2%) (full text).