Sleep Low carbohydrate periodization: ~3% faster 10-km run, better cycling economy in 3 weeks
Participants
n=21 trained triathletes, VO2max 58.7 ± 5.7 ml/kg/min: Sleep-Low (n=11) vs Control (n=10)
Protocol (3 weeks)
Equal intake: both groups 6 g/kg/day carbohydrate, differing only in timing
Sleep-Low: three blocks of (1) "train-high" evening interval session with high CHO availability → (2) overnight CHO restriction ("sleeping low") → (3) "train-low" session with low endogenous and exogenous CHO availability
Control: same training, high CHO availability throughout (no overnight restriction, CHO provided during sessions)
Key Results
Delta efficiency, submaximal cycling: SL +11 ± 15% vs CON +1.4 ± 9.3% (p<0.05)
Supramaximal cycling to exhaustion at 150% of peak aerobic power: SL +12.5 ± 19.0% vs CON +1.63 ± 12.4% — p=0.06, a trend, not significant
10-km running performance: SL −2.9 ± 2.15% time vs CON −0.10 ± 2.03% (p<0.05)
Fat mass: SL −8.5 ± 7.4% vs CON −2.6 ± 7.4% (p<0.01); lean mass unchanged (SL −0.16 ± 1.7%, CON −0.22 ± 1.0%)
Critical Nuance
Total carbohydrate intake was identical (6 g/kg/day). The only difference is timing around sessions — periodization of availability, not a deficit.
Practical Application
Finding: in 21 trained triathletes, 3 weeks of sleep-low periodization (evening intervals with carbs → no carbs overnight → low-carb morning session) at the same 6 g/kg/day improved 10-km running time by 2.9% vs 0.1% in controls and cycling delta efficiency by 11% vs 1.4%; fat mass fell 8.5% with lean mass preserved. The supramaximal-capacity gain (150% PAP) was only a trend (p=0.06).
How it was done: the low-carb session followed a hard evening session with carbs — the quality work is fuelled, only the next-morning session is done low.
Not from this paper; the base's own practice: keep sleep-low to a few blocks per week for short periods, never around key high-intensity sessions, and fuel fully in test/race weeks — low availability combined with heavy training raises the RED-S / under-recovery risk (see REF-349).
Limitations
n=21, two parallel groups, 3 weeks; randomized (stated in the full text, not in the abstract) (corrected 2026-10-12, full text); trained triathletes only
Supramaximal-capacity gain was a trend (p=0.06); the 10-km test was running, the economy test cycling — no cycling time-trial outcome
How this entry was checked and changed
Why this level
regraded 2026-09-27 from A: parallel randomized trial, n=21
Content correction 2026-09-27
Supramaximal capacity "↑ (p<0.05)" -> p=0.06 (trend, not significant per abstract); fat-mass p<0.05 -> p<0.01; the 10-km TT is a running test, now stated.
Practical "2 blocks/week × 2 weeks = 4 blocks" -> not in the abstract (the study used a 3-wk intervention of three block types); removed, replaced by a labelled house-practice line. "Burke 2015 — RED-S risk" reference without refId removed.
Type "RCT" -> two parallel groups ("divided into two groups", randomization not stated in the abstract; the full text says "randomly divided into two groups" (corrected 2026-10-12, full text)); level A -> B+ (parallel controlled trial, n=21). Sport line adds running (10-km run was the performance test).
Full text (checked 2026-10-12)
Correction: the entry says the groups were "not described as randomized" / "randomization not stated"; the paper says "The study used a parallel group design, with the subject cohort being randomly divided into two groups" and participants "were randomly assigned" to SL (n=11) or CON (n=10) (full text).
Participants were 21 endurance-trained male triathletes who had been competing for > 2 years and were training a minimum of 10 hours per week (full text).
The intervention block was six sessions over four consecutive days: HIT in the afternoon (alternately 8 x 5 min cycling at 85% MAP or 6x5 min running at individual 10 km intensity, 1 min recovery between sets) and LIT the next morning (60 min cycling at 65% MAP), with one LIT session per day on the other days for a total of 10-15 h (full text).
SL took no CHO from immediately after HIT until completion of the next-morning LIT, none during HIT and LIT sessions, ate a CHO-free evening meal and did LIT after an overnight fast; CON drank a 6% CHO-electrolyte drink during sessions; both groups had a high protein sugar-free drink (15 g per 20 mL) before bed (full text).
Reported intake during the intervention was 5.44 ± 1.2 g·kg-1·d-1 CHO for SL (dinner: 0.0 ± 0.1 g·kg-1) and 5.65 ± 0.99 g·kg-1·d-1 CHO for CON (full text).
10 km running time in SL went from 40:23 ± 03:22 to 39:10 ± 03:02 min:s (-2.9 ± 2.1%, P < 0.01, d=0.38), and supramaximal time to exhaustion from 52.7 ± 13.8 s to 57.8 ± 6.4 s (P= 0.062, d=0.41) (full text).
RPE during the LIT sessions (fasted for SL) was higher in SL than CON: 14.3 ± 0.29 vs. 12.7 ± 0.43 (P < 0.05) (full text).
SL body mass fell from 70.6 ± 5.0 kg to 69.6 ± 5.0 kg (P < 0.001, d=0.19) and fat mass from 9.70 ± 4.08 kg to 8.86 ± 3.8 kg, with no change in CON and no change in fat free mass in either group (full text).