REF-706 · Wehrlin & Hallén, 2006 · European Journal of Applied Physiology, 2006 Mar;96 · Paper · PubMed
Aerobic performance declines LINEARLY with altitude
About −6.3% VO₂max per 1000 m — and the loss is already statistically significant at 800 m: there is no threshold below which altitude is free
Protocol
n=8 endurance-trained athletes, tested running. In randomised double-blind order, on separate days, each ran to exhaustion in a hypobaric chamber at six simulated altitudes: 300, 800, 1300, 1800, 2300 and 2800 m. 300 m served as the sea-level reference.
The key design feature: the test was run at the same absolute speed at every altitude — 107% of the speed associated with sea-level VO₂max. This removes the standard objection to earlier work ("VO₂max is lower at altitude simply because the athlete cannot reach the same intensity"): here the intensity was identical by construction.
VO₂max = highest oxygen uptake in that test; performance = time to exhaustion. Douglas bags for respiratory measures, pulse oximetry for SpO₂.
Key Findings
| Altitude | VO₂max (ml·kg⁻¹·min⁻¹) | SpO₂ at test end |
|---|---|---|
| 300 m | 66 ± 1.6 | 89.0 ± 2.9% |
| 2800 m | 55 ± 1.6 | 76.5 ± 4.0% |
Slope: −6.3% VO₂max per 1000 m of altitude gain. Individual range 4.6-7.5% — a 1.6-fold spread between the most and least affected athlete.
Time to exhaustion at fixed speed fell 14.5% per 1000 m (p<0.001). The performance loss is about 2.3 times the VO₂max loss (14.5 ÷ 6.3). Inference, not tested: the test was above threshold, where a small shift in capacity changes time to exhaustion sharply.
The decline is linear across the whole 300→2800 m range. There is no breakpoint. The drop between 300 and 1300 m is essentially the same as between 800 and 1800 m.
Already at 800 m both VO₂max (P<0.01) and performance (P<0.05) were significantly below the 300 m values.
Inference, not tested: the convention that "altitude starts" well above this is a classification threshold, not a physiological one.
Practical Application
In n=8 endurance-trained athletes on acute exposure, VO₂max fell linearly by 6.3% per 1000 m (individual range 4.6-7.5%) from 300 to 2800 m, and was already lower at 800 m — there is no altitude threshold below which performance is free. Working rule for an unacclimatized athlete: about −6% VO₂max per 1000 m gained.
Inference, not tested (linear arithmetic on the 6.3%/1000 m slope, valid only within the tested 300-2800 m): +1000 m ≈ −6%, +1500 m ≈ −9%, +2000 m ≈ −13%, +2500 m ≈ −16%.
Not from this paper; the base's own practice — reviewing a session tagged `altitude`: lower power at the same RPE and heart rate is expected physiology, not lost fitness and not poor execution. Do not re-estimate FTP downward from a mountain session and do not log it as a decline in the trend.
Not from this paper; the base's own practice — set the expectation before the trip, do not explain it afterwards. An athlete heading to a camp or race at moderate altitude will lose noticeable power, and without warning will read that as detraining.
Moderate altitude is not zero. Performance was already lower at 800 m in this study. Inference, not tested: acclimatization may be worth considering for moderate-altitude races too.
Individualise from the athlete's own history. The 4.6-7.5% spread means the group rule is a starting hypothesis. Not from this paper; the base's own practice: with two or more activities at similar altitude, that athlete's own power offset beats the table.
Limitations
n=8, running, acute exposure, single laboratory, 2006. There was no acclimatization at all — the slope describes a first encounter with altitude, not the state after 1-2 weeks.
The −14.5% per 1000 m figure applies to time to exhaustion at a fixed supra-threshold speed (107% of sea-level VO₂max speed). It does NOT mean a 40 km time trial slows by 14.5% per 1000 m. That number is specific to the severe domain and sensitive to test duration. The transferable quantity here is the VO₂max slope.
Upper bound is 2800 m. Linearity above that was not tested.
Hypobaric chamber, not real mountains: no cold, wind, dehydration, disrupted sleep or accumulated travel. Inference, not tested: in the field the losses may be larger.
Threshold and lactate-threshold power were not measured directly — only VO₂max and time to exhaustion.
How this entry was checked and changed
Why this level
regraded 2026-09-27 from B+: crossover, n=8
Content correction 2026-09-27
Level B+ -> B (crossover, n=8, per rubric). Removed numbers not in the abstract: participant VO₂max range 61-74, age ~24, times to exhaustion ~5:20 and ~3:30; added the abstract's SpO₂ values.
"Two and a half times" -> 2.3× (14.5 ÷ 6.3); altitude ladder corrected to the slope's arithmetic (2000 m gain ≈ −13%, 2500 m ≈ −16%) and labelled as inference.
Removed the "Refines REF-210" bullet (REF-210 does not state a 10-15% LT loss at 2500 m) and "other work suggests a steeper decline" (unsourced); fixed REF-209 / REF-199 cross-reference blurbs to what those entries say.
FTP/trend-review and pre-trip advice labelled as the base's own practice; lead ▶ now states the study's finding and population.