The Cognitive Costs of Sleep Deprivation in Ultra-Endurance Cycling: Insights From the Race Across France
Twenty-three riders in the Race Across France wore a wrist accelerometer from before the start to the finish, rated their sleepiness every 4 h and did a cognitive task at basecamps five times (before the start, at 620, 1438 and 2066 km, and after the finish). Riders with more sleep per 24 h finished higher in the rankings.
Sleepiness was greater in those who slept less than 5.29 h a day on average and followed a distinct 24 h rhythm, and cognitive performance was worse and response times slower in those who slept less across days.
The authors conclude that sleep duration and timing are imperative for physical and cognitive well-being in ultracycling and that aligning sleep with homeostatic and circadian drives may improve performance.
Participants
n=23 ultracyclists who completed the study (abstract: "Twenty-three cyclists completed the study"). Sex, age, experience and finishing times are not given in the abstract.
Protocol
Wrist-worn accelerometer (actigraphy) worn throughout the race to track sleep.
Sleepiness rated every 4 h from before the start to the end.
Cognitive task at basecamp on five occasions: before the start, at 620 km, 1438 km and 2066 km, and after the finish.
Analyses: sleep strategy and duration vs race time/ranking; sleep debt and circadian timing vs sleepiness and cognitive performance.
Key Results
Higher mean sleep time per 24 h went with higher race rankings.
Perceived sleepiness was greater in riders who slept less than 5.29 h per day on average, with a distinct 24 h rhythm.
Cognitive performance was worse and response times slower in riders who slept less across days.
The abstract gives no effect sizes, correlation coefficients or P values.
Practical Application
In 23 riders tracked with wrist actigraphy through the Race Across France, those who slept more per 24 h finished higher, those averaging under 5.29 h a day were sleepier, and less sleep across days went with worse cognitive performance and slower responses. To "should I cut sleep to gain time in a multi-day race?": in this race, riders who slept less did not finish higher, and they paid for it in alertness.
The 5.29 h figure is where the sleepiness comparison was split, not a tested minimum dose; do not hand it to a rider as a target.
The authors recommend timing sleep with the homeostatic and circadian drives for sleep; they did not test a sleep plan as an intervention. Inference, not tested here: in practice that means placing the main sleep stops in the night hours, when both drives are high, rather than taking scattered naps at random times.
Slower responses on the bike are a crash risk. Not from this paper; the base's own practice: treat disorientation or hallucinations in an ultra as a stop-and-assess signal (REF-665).
Limitations
Correlational: faster, fitter or more experienced riders may both sleep more and finish higher; the design cannot separate cause from association.
n=23 finishers of one race; sex, age and level are not in the abstract.
Sleep measured by wrist actigraphy (an estimate, not polysomnography); cognition tested only at basecamps, five times.
No effect sizes in the abstract. Written from the abstract and record metadata only (full text, open access, not read).
How this entry was checked and changed
Why this level
observational cohort, n=23; associations only
Full text (checked 2026-10-12)
The June 2024 race covered 2588 km with 29,250 m of positive elevation within a time limit of 240 h, self-supported, with riders planning their own sleep; the investigators did not control sleep/wake patterns (full text).
Of 40 volunteers (including five women), 23 (two women) completed the protocol, mean age 37.2 ± 9.1 years; 10 abandoned the race and seven were discarded for protocol failures (full text, Table 1).
Participants ranked 3rd to 96th of the 134 finishers, were on course for 142.1 to 214.2 h (190.7 ± 12.41 h) and spent 5–8 nights in the race (full text).
Actigraphic total sleep time was 228 ± 107 min per 24 h (individual means 95 to 306 min), and race ranking was related to mean total sleep time per 24 h with r = 0.80 (p < 0.001) (full text).
Sleep propensity was highest between 0:00 and 4:00 with no secondary peak; sleep made up 39.2% ± 6.7% of GPS null-velocity time, and this ratio was not correlated with ranking (full text).
Sleepiness peaked between 23 and 00 h; response times were slowest from baseline between 5 and 9 h and fastest between 17 and 21 h, and less sleep between tests was associated with slower response times (F = 31.51, p < 0.001) (full text).
Cognition was measured with a Go/No-Go Reversal task (56 trials before and 56 after the reversal), and the majority of cognitive tests were completed in the early evening, which the authors note could have influenced the results (full text).