Same rider, same watts, same fluid replacement: heart rate drifted 11% and VO2max fell 15% at 35 °C, while at 22 °C it drifted 2% and VO2max did not significantly change
The environment, not the athlete, wrote most of the number
Participants
n=10 endurance-trained men, 23 ± 3 y, VO2max 64.7 ± 8.7 mL/kg/min.
Protocol
45 min of cycling at 59.2 ± 1.9% VO2max in a hot condition (35 °C) and a cool one (22 °C) on separate days. Cardiovascular drift was measured as the change in heart rate and stroke volume between minute 15 and minute 45.
VO2max was measured immediately after the 45 min, and separately after only 15 min of cycling in both conditions — so the fall in VO2max is assessed across exactly the window in which the drift occurred. Fluid was given to replace sweat losses in all trials.
Key Findings
In the heat (35 °C), between minutes 15 and 45: heart rate +11%, stroke volume −11%, VO2max −15% (all p<0.05).
In the cool (22 °C), across the same window: heart rate +2%, stroke volume −2%, and the 5% fall in VO2max was NOT statistically significant (p>0.05).
Both the drift and the accompanying fall in maximal oxygen uptake were significantly greater in the heat than in the cool (p<0.05).
Dehydration was not the cause. Sweat losses were replaced in every trial, and the heat effect appeared anyway — so a large drift on a hot day is not evidence that the athlete drank too little.
The magnitudes travel together: an 11% rise in heart rate came with an 11% fall in stroke volume and a 15% fall in the ceiling. Drift is a redistribution, not merely an inconvenience.
Practical Application
Anchor the reading in ambient temperature before anything else. At a genuinely moderate effort with fluid replaced, a cool-weather session drifts about 2% over half an hour and a hot one about 11%. A large drift on a hot ride is the expected result, not a finding about the athlete's fitness.
Do not diagnose dehydration from drift alone. Sweat losses were fully replaced here and the heat still produced the effect. Under-drinking adds to drift; it is not what creates it.
Compare an athlete's drift only against their own sessions in similar conditions. Cool against cool, hot against hot. A ride-to-ride comparison across a 13 °C temperature difference (the 35 °C vs 22 °C gap tested here) is comparing environments.
Judge drift over a continuous, steady, moderate effort — here the window was minutes 15 to 45 of a 45-min ride at 59.2% VO2max. That is the measurement this experiment describes; a session of repeated intervals with recoveries does not produce it.
Expect the ceiling to fall with the drift. VO2max measured right after the 45 min fell 15% in the heat. Inference, not tested: the late part of a hot race is being ridden against a smaller maximum, which is a pacing instruction, not a mindset problem.
Limitations
n=10, young, well-trained men. Not women, not masters athletes, not recreational riders — and a 64.7 mL/kg/min group is fitter than most of this app's population.
Written from the journal's complete structured abstract (purpose, methods, results, conclusion). The full text is paywalled and was not read, so figures, time-course detail and secondary measures are not reproduced here.
One intensity (59.2% VO2max) and one duration (45 min). It does not establish how drift scales at threshold intensity or over four hours.
The 5% and 10% decoupling thresholds used in this app are house conventions and are NOT from this paper. The 2% and 11% here are condition means over a 30-min window at a fixed submaximal load — they are not a pass/fail line, and they should not be quoted as one.
Glycogen depletion is a known confounder of drift and was not manipulated here; a 45-min ride does not test it.
Cycling only, laboratory only, with fluid replacement that few athletes achieve in the field.
How this entry was checked and changed
Content correction 2026-09-27
"at least 30-45 min" -> replaced with the measured window (minutes 15-45 of a 45-min ride at 59.2% VO2max); "13 °C" stated as the 35 vs 22 °C gap; "~59%" -> 59.2%.
"final hour of a hot race" -> relabelled inference; the study measured 45 min.
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).