REF-708 · Koistinen, Takala, Martikkala & Leppäluoto, 1995 · International Journal of Sports Medicine, 1995 Feb;16 · Paper · PubMed
The spread in altitude losses is not random: the higher the sea-level VO₂max and lactate threshold, the LARGER the decline at a simulated 3000 m
The fittest athletes lose the most
Protocol
n=12 highly trained athletes — 6 male ice-hockey players and 6 cross-country skiers (2 women and 4 men among the skiers). Each performed a maximal incremental cycle-ergometer test twice, on separate days:
1
at sea level (normobaric conditions);
2
in a hypobaric chamber simulating 3000 m (520 mmHg).
The question: does the decline in VO₂max and in the lactate threshold at altitude depend on how fit the athlete is at sea level?
Key Findings
| Measure | Sea level | 3000 m | Mean loss |
|---|---|---|---|
| VO₂max | 57.4 ± 7.1 | 46.6 ± 4.9 | ≈ −19% |
| Lactate threshold (VO₂ at LT) | 43.5 ± 6.4 | 36.5 ± 4.2 | ≈ −16% |
The spread is partly explained by sea-level fitness. The higher the sea-level VO₂max, the larger its decrease at 3000 m: r = 0.61 (p<0.05), i.e. about 37% of the between-athlete variance (r²). (Mean losses in the table are arithmetic on the abstract's means.)
For the lactate threshold: r = 0.68 (p<0.01), about 46% of the variance (r²). Inference, not tested: the two correlations were not formally compared, so "the threshold is more fitness-dependent" is not established.
Inference, not tested here (proposed mechanism): highly trained athletes have a larger maximal stroke volume and cardiac output, so blood transits the pulmonary capillaries faster; at reduced oxygen partial pressure haemoglobin does not fully saturate (exercise-induced arterial hypoxaemia, EIAH), lowering arterial oxygen content and delivery.
Practical Application
In n=12 highly trained athletes at a simulated 3000 m, VO₂max fell from 57.4 to 46.6 and lactate-threshold VO₂ from 43.5 to 36.5 ml·kg⁻¹·min⁻¹, and the fittest athletes lost the most (r = 0.61 for VO₂max, r = 0.68 for the threshold) — the individual spread in altitude losses is partly predictable from sea-level fitness. It complements REF-706, which gives the group slope and an individual range of 4.6-7.5% (Not from this paper: REF-706).
Inference, not tested — do not apply the group rule uniformly across a group. On a mountain trip the strongest rider loses the most in percentage terms — the field compresses. That is expected, not evidence that they have lost form or prepared badly.
Inference, not tested — give acclimatization priority to the strongest athletes rather than treating it as an elite luxury. Those with the most to lose may have the most to regain.
Inference, not tested — when setting expectations for a recreational athlete, do not quote elite numbers. They will likely lose less than a better-trained athlete beside them; recreational athletes were not studied here.
The lactate threshold also fell at 3000 m (43.5 -> 36.5 ml·kg⁻¹·min⁻¹), and more so in the fitter athletes — so training zones, not only VO₂max, shift at altitude.
Limitations
n=12, published 1995, mixed sample (ice-hockey players plus skiers, men and women pooled). Very small.
A single altitude — 3000 m. Whether the fitness dependence holds at lower, more common race altitudes is not addressed.
Statistical caveat (inference): correlating a baseline value with the change from that same baseline is mathematically predisposed to inflation (regression to the mean, part-whole coupling). The direction of the effect is supported by a plausible mechanism, but the magnitude (37% / 46% of variance) should be read as an upper estimate.
r = 0.61 and 0.68 at n=12 carry very wide confidence intervals. Even at face value, more than half the spread remains unexplained: fitness is neither the only nor the dominant predictor.
EIAH was not measured here. No saturation, blood-gas or transit-time data — the explanation is imported from adjacent literature, not demonstrated by these data.
Sample mean VO₂max 57.4 is "trained", not world class. Extension to professionals remains an inference by analogy.
How this entry was checked and changed
Content correction 2026-09-27
Removed "individual range 0% to −30%" (not in the abstract). "Percentage decline" -> decline: the abstract correlates the decrease (delta) with sea-level values and does not say percentage; "first lactate threshold (LT1)" -> lactate threshold.
"The threshold is more fitness-dependent than VO₂max" relabelled as untested (the correlations were not compared); acclimatization-priority and recreational-expectation advice labelled as inference; unsourced "1000-1500 m" and "VO₂max 75-85" removed.
Lead ▶ now states the study's numbers and population; REF-210 cross-reference blurb corrected (it covers altitude training for sea-level performance, not acclimatization).
Sport line -> "cycling" (the test modality; the ice-hockey and cross-country-ski sample is described under Protocol).