REF-1365 · Goulet, 2011 · British Journal of Sports Medicine · Paper · PubMed

Will losing 2% of my body weight in sweat slow me down?

Five cycling time trials near 26 °C: no power cost from a 2.2% loss; thirst beat under-drinking. A blinded trial at 34 °C: 8% less work (REF-226). Unsettled (B+, abstract).

Main caveat: Abstract only. The full text is behind the publisher's paywall, with no Europe PMC copy and no repository copy listed by OpenAlex.
moderate evidenceMeasured on recreational to highly trained athletes

Goulet set out to measure how large the effect of exercise-induced dehydration (EID) is on time-trial performance, the test in which athletes pace themselves as in a race. Five research articles met his criteria, all using cycling time trials, giving 13 effect estimates from 39 subjects.

The trials averaged 26.0 ± 6.7 °C, 61 ± 9% relative humidity, 68 ± 14% of VO2max and 86 ± 34 min. Performance was expressed as the percentage change in power output. Against euhydration (mean body-mass loss 0.44 ± 0.48%), EID with a mean body-mass loss of 2.20 ± 1.0% during self-paced exercise changed endurance performance by a non-significant +0.06 ± 2.72% (p = 0.94).

Meta-regression found that the change in power output was related to exercise intensity and duration, not to the percentage of body mass lost.

Drinking according to thirst went with better time trials than drinking below thirst (+5.2 ± 4.6%, p = 0.01) and, not significantly, than drinking above it (+2.4 ± 5.0%, p = 0.40); the probability that thirst drinking gives a real, meaningful advantage under field conditions was put at 98% and 62% against those two strategies.

The author's three conclusions: compared with euhydration, EID up to 4% of body mass does not alter cycling performance during out-of-door exercise conditions; intensity and duration matter far more than EID; and relying on thirst maximises cycling time-trial performance.

Our reading: this is the base's named dissent on the "2% costs performance" line in PHYS-30 and SYNTH-68 — a small pooled analysis in which a typical race-like loss of about 2% cost nothing in self-paced cycling, and in which drinking to thirst came out best.

It is set against blinded evidence that about 2.4% hypohydration cut work by about 8% at 34 °C (REF-226, seven men) and a position that losses above 2% usually impair aerobic performance in the warm (REF-062). Two different things are concluded here and should not be merged: no effect of losses up to 4% against euhydration, and thirst drinking as the best strategy.

Protocol

Design: meta-analysis. Studies located via database searches and cross-referencing; time-trial outcomes converted to mean percentage changes in power output; random-effects meta-regression, an ANOVA-analogue and weighted mean effect summaries to relate body-mass loss to time-trial performance.
Sample: n = 39 subjects in 5 studies (13 effect estimates).
Conditions (means ± SD across trials): ambient temperature 26.0 ± 6.7 °C; relative humidity 61 ± 9%; exercise intensity 68 ± 14% of VO2max; duration 86 ± 34 min.
Comparisons: EID (mean body-mass loss 2.20 ± 1.0%) vs euhydration (0.44 ± 0.48%); drinking to thirst vs drinking below or above thirst.
Not given in the abstract: the five studies and their participants; how hydration was manipulated in each and whether participants knew their hydration state; whether the time trials were run outdoors or in a laboratory, and with what airflow; the number of effect estimates above 3% loss; a risk-of-bias appraisal; funding and competing interests.

Key Results

Outcome (abstract)Result
Pool5 articles, all cycling time trials; 13 effect estimates; 39 subjects
Body-mass lossEID 2.20 ± 1.0%; euhydration 0.44 ± 0.48%
EID vs euhydration, self-paced exercise+0.06 ± 2.72% power output, p = 0.94 (not significant)
Meta-regressionpower change related to exercise intensity and duration (significant), not to the EID-related % body-mass loss
Drinking to thirst vs below thirst+5.2 ± 4.6%, p = 0.01
Drinking to thirst vs above thirst+2.4 ± 5.0%, p = 0.40
Probability of a real, meaningful advantage of thirst drinking under field conditionsabout 98% vs below thirst; about 62% vs above thirst
Conclusion 1EID up to 4% body-mass loss does not alter cycling performance during out-of-door exercise conditions
Conclusion 2intensity and duration have a much greater impact on cycling time trials than EID
Conclusion 3relying on thirst to gauge fluid replacement maximises cycling time-trial performance

Mechanisms

Measured (pooled): time-trial power did not fall with an average 2.2% loss of body mass in self-paced cycling; the size of the loss did not predict the change in power, while intensity and duration did.
Inference (from the wording of the abstract): the author limits his first conclusion to out-of-door exercise conditions and the thirst probability to field conditions, which suggests he does not regard laboratory conditions as equivalent; the abstract gives no reason (for example airflow and convective cooling), and none is claimed here.
Inference: time trials allow pacing, so a dehydrated rider may slow early and gain back little; the measure is the race-like outcome, not physiological strain, which may still rise (heart rate and core temperature were not reported in the abstract).
Inference: the base's blinded trial (REF-226) removed expectation and was run at 34 °C after an intermittent preload of 8 × 15 min, hotter than this pool's mean; the two sets of results can both be true for their own conditions.

Practical Application

Will losing 2% of my body weight in sweat slow me down? Five cycling time trials near 26 °C: no power cost from a 2.2% loss; thirst beat under-drinking. A blinded trial at 34 °C: 8% less work (REF-226). Unsettled (B+, abstract).
What it is good for as a source: naming the dissent on the 2% threshold with its exact scope: self-paced cycling, about 26 °C and 86 min, five studies, 39 subjects, and a conclusion worded for outdoor conditions.
What it does not show: that dehydration never matters (heat, longer events and running are outside the pool), or that losses up to 4% were tested often (the mean was 2.2%). "No effect up to 4%" and "thirst drinking is best" are two separate conclusions.
How a principle should cite it: "a meta-analysis of five cycling time-trial studies found no performance loss with an average 2.2% loss (the author concludes up to 4% in outdoor conditions), and drinking to thirst performed best (Goulet 2011, from the abstract)".
What it means for training: Not from this paper; the base's own practice: drink to thirst by default, with a weighed sweat rate as a ceiling, never a target (SYNTH-68, REF-776); in hot, long events a large deficit is still treated as a cost (REF-062, REF-226).

Limitations

Abstract only. The full text is behind the publisher's paywall, with no Europe PMC copy and no repository copy listed by OpenAlex. The five studies, their participants, how hydration was set, whether subjects were blinded, where the trials were run and with what airflow, any risk-of-bias appraisal, and the funding and competing interests could not be read.
Small pool. Five studies, 39 subjects, all cycling; the confidence in a null pooled effect is limited by its spread (± 2.72%).
Thirst comparisons are not described. The abstract does not say whether thirst vs below or above thirst was compared within trials or across them, or how many effect estimates each comparison used; "associated with" is its wording.
Later criticism is not covered here. The abstract gives none, and no later paper's claims are added; the base's opposing evidence is REF-226 and REF-062.
Single author. One reviewer for search, extraction and analysis, as far as the abstract shows.
How this entry was checked and changed
Why this level

a meta-analysis, which the rubric places at A- when the trials are small; held at B+ because the pool is five studies and 39 subjects in all (less than one well-powered trial), every study is a cycling time trial, no risk-of-bias appraisal is reported in the abstract, and the full text was not read (paywalled; no open copy)