REF-1204 · Jones et al., 2026 · Scandinavian Journal of Medicine & Science in Sports · Paper · PubMed · Free full text
How much carbohydrate, and for how long, to load before a race?
A review of 63 biopsy trials advises over 8 g/kg a day for 36-48 h; the amount mattered most, the best duration is unsettled. Glycogen, not race time; pays past ~90 min.
Jones et al. pooled the biopsy studies that reported carbohydrate intake and exercise for at least 24 h before a muscle glycogen measurement, to find out how much carbohydrate, for how long, and with what exercise maximises glycogen before competition.
The descriptive synthesis covered 63 trials in 61 papers (641 participants, 12% women); 18 randomised trials comparing under 6.5 with over 6.5 g/kg/day at matched duration entered the meta-analysis.
Going from the lower to the higher intake raised glycogen by 191.9 mmol/kg dry mass in the 13 crossover-type trials and by a non-significant 107.6 in the 5 parallel-group trials, with heterogeneity above 90% in both.
Exploratory regressions showed glycogen rising linearly with intake in recreationally active (r² = 0.332) and endurance-trained groups (r² = 0.490), flattening only slightly above 10 g/kg/day; in the well-trained intake did not predict glycogen (r² = 0.086, p = 0.147).
In pooled subgroups 24 and 48 h of loading raised glycogen more than 72 and 96 h (subgroup difference p = 0.01), but the 24-h subgroup had 4 participants and the subgroups differed in how much extra carbohydrate was eaten, so the authors say an optimal duration cannot be determined from the analysis.
Their practical rule, built from the pooled and single studies together, is more than 8 g/kg/day for 36-48 h, or 24-36 h when prior glycogen-depleting (or high-intensity) exercise and high-glycaemic carbohydrate start it.
Our reading: the best available synthesis on loading dose and timing, and it is consistent with the 1-2 days at 10-12 g/kg in REF-748 but cannot show that longer loading is worse; it is about glycogen in the thigh, not about race time, and its duration and training-status subgroups are too small and confounded to carry exact numbers.
Protocol
Registration and reporting: Open Science Framework, December 2023; PRISMA and the Cochrane handbook. Searches of PubMed, Web of Science, Cochrane, SportDiscus and Scopus by two researchers in February 2022 and December 2024, updated June 2026; citation searching added 16 studies.
Inclusion: healthy adults aged 18-60; carbohydrate intake in g/kg/day (or calculable) and exercise status reported for ≥ 24 h before the biopsy; whole-muscle glycogen by biopsy in dry mass (or convertible); randomised, counterbalanced or non-randomised controlled trials. Intake categories: low < 3, moderate 3-6.5, high > 6.5-9, very high > 9 g/kg/day.
Data: values shown only in figures (17 studies) were digitised; wet-mass values converted to dry mass with published factors. Training status graded 1-5 (sedentary to elite).
Exploratory regression: simple linear and second-degree polynomial fits of intake against glycogen by training status; observations were not independent (multiple time points per study), handled in sensitivity analyses.
Meta-analysis: randomised trials with no more than one risk-of-bias domain of "some concerns", comparing a low-moderate (0-6.5) with a high-very high (> 6.5 g/kg/day) intake held constant over a matched loading period; generic inverse-variance random effects (RevMan 5.4.1); crossover and parallel designs analysed separately, with correlations borrowed for 7 crossover trials.Subgroups (crossover trials only): difference in relative intake (1-g steps), difference in absolute intake (100-g steps), loading duration and training status; sex and biopsy site not analysed because of uneven distributions. Heterogeneity significant at p < 0.10.
Risk of bias: RoB 2 for randomised and ROBINS-I for non-randomised trials, two reviewers independently.
Funding and interests: the funding statement reads "The authors have nothing to report"; no conflicts of interest declared.
Key Results
Meta-analysis, low-moderate (< 6.5) vs high-very high (> 6.5 g/kg/day) carbohydrate, glycogen in mmol/kg dry mass (DM):
| Analysis | Comparisons (participants) | Mean extra intake (g/kg/day) | Mean difference [95% CI] | Heterogeneity I² |
|---|---|---|---|---|
| Crossover-type trials, overall | 13 studies (111) | — | +191.9 [142.8, 241.1], p < 0.0001 | 94% |
| Parallel-group trials, overall | 5 studies (81) | — | +107.6 [−10.5, 225.7], p = 0.07 | 92% |
| Loading 24 h | 2 (4) | 3.1 | +207.7 [147.0, 268.4] | 0% |
| Loading 48 h | 7 (56) | 6.0 | +255.0 [160.7, 349.2] | 97% |
| Loading 72 h | 5 (38) | 4.1 | +148.0 [90.0, 206.0] | 85% |
| Loading 96 h | 3 (13) | 3.4 | +106.4 [59.0, 153.7] | 40% |
| Trained participants | 10 (47) | 4.9 | +229.9 [147.1, 312.6] | 94% |
| Well-trained participants | 6 (42) | 4.0 | +136.6 [86.5, 186.8] | 88% |
Duration subgroups differed (p = 0.01); training-status subgroups did not (p = 0.06).
Dose: the extra relative intake raised glycogen stepwise across subgroups (p = 0.01) with a possible flattening above 5.5 g/kg/day extra, but every subgroup stayed heterogeneous (I² 54-96%). By extra absolute intake the rise continued (p < 0.0001), the > 500 g/day subgroup (2 comparisons) gaining the most, +401.3 mmol/kg DM.
Regression: relative intake explained 33.2% of glycogen variation in recreationally active groups (p < 0.001), 49.0% in endurance-trained groups (p < 0.001) and 8.6% in well-trained groups (p = 0.147); only five study groups ate about 12 g/kg/day.
Training status (single studies): after about 72 h of very high intake trained participants stored more than untrained (583.3 vs 287.1 and 792.1 vs 584.6 mmol/kg DM), and 10 weeks of cycling training raised the same people's loaded glycogen from 428.5 to 795.6.
Exercise: 50 of 63 trials exercised participants before loading; the same paragraph then counts 44 of 48 as using a specific depleting protocol (two different denominators in the original). Only 22 (35%) allowed exercise during loading, mostly light-to-moderate sessions of 60 min or less, which did not appear to lower glycogen when intake was very high (between-study comparison).
Women: 74 of 641 participants; 79% of experiments were men only and only 5% women only. When intake per kilogram or in grams was matched, women stored as much as men (descriptive; REF-1205, REF-1206 and Walker 2000 among the studies). The review attributes the sex difference in Tarnopolsky's 1995 study to intake (6.4 vs 8.2 g/kg/day in women and men).At 8-10 g/kg/day no menstrual-phase difference appeared; at moderate intakes some studies found more glycogen in the mid-luteal phase.
Carbohydrate type: 28 of 63 trials did not describe it well enough to judge glycaemic index; 5 fed only high-glycaemic sources; 2 compared types.
Diet control: 84% of trials supplied the food, but only 10% of those measured compliance.
Risk of bias: 24 of 42 randomised trials had some concerns (mostly unclear randomisation and allocation concealment), 18 were low risk; 13 of 21 non-randomised trials had a serious risk of bias from confounding.
Where the abstract and the body differ: the abstract quotes only the crossover estimate, not the non-significant parallel-group one. Its "> 8 g/kg/day for 36-48 h" is the authors' practical synthesis; the pooled threshold was 6.5 g/kg/day, and the Discussion calls the duration estimates untrustworthy and says an optimal loading duration cannot be determined from this analysis.
More numbers that do not add up: the Methods count 12 crossover trials, the Results 13 repeated-measures trials (a mixed-design trial treated as crossover is the likely difference); and the training-status paragraph says "an additional 24 participants" contributed to the endurance-trained subgroup, while Table 2 gives 47 vs 42.
A study in the base re-read: the review says the "1-day" protocol of Bussau 2002 (REF-891) was really about 36 h of loading, because the last habitual training session came about 12 h before it.
Mechanisms
Measured: glycogen rose with intake in recreationally active and endurance-trained groups; trained people stored more than untrained ones at the same high intake; light exercise during loading did not lower stores when intake was very high (between studies).
Authors' interpretation: endurance training raises GLUT4 and glycogen synthase activity and the share of type I fibres, so trained muscle stores more and may need less extra carbohydrate to fill — one possible reason intake did not predict glycogen in the well-trained (the other being few study groups at the extremes).Exercise before loading drives local glycogen synthesis; high-glycaemic carbohydrate speeds storage over short windows (< 24-36 h), while over longer windows the type matters little.
Authors' interpretation (untested sequence): a final hard session 2-3 days out, then 1-1.2 g/kg/h of carbohydrate for the first 4 h and a high to very high intake (above 6.5 g/kg/day; their summary rule elsewhere is > 8) for 36-48 h without further moderate-to-hard exercise, should in theory maximise stores. The authors label it theory; the 2-3-day timing of elite athletes' last hard session is their unpublished field observation.
Inference: the loading-duration subgroups mix duration with dose (the 48-h trials added 6.0 g/kg/day, the 72-h trials 4.1), so "24-48 h beats 72-96 h" is as much a statement about how much extra was eaten.
Practical Application
How much carbohydrate, and for how long, to load before a race? A review of 63 biopsy trials advises over 8 g/kg a day for 36-48 h; the amount mattered most, the best duration is unsettled. Glycogen, not race time; pays past ~90 min.
Dose before duration: the amount eaten was the main lever (A-, glycogen outcome).This agrees with REF-748 (A, 10-12 g/kg/day for 1-2 days before a single-day event) and is compatible with the 2-3-day high-carbohydrate phase quoted in SYNTH-65 (a practice line from the REF-774 entry) and with SYNTH-64's practice (1-3 days, 10 g/kg): the authors' rule is 36-48 h, but their duration subgroups are confounded by dose and they say an optimal duration cannot be determined, so a third day is neither shown to be needed nor shown to harm.
Short on time (24-36 h): start right after a hard or glycogen-depleting session and use high-glycaemic carbohydrate (the authors' rule from single studies; REF-891: 10 g/kg/day at rest, B).
Keep the easy taper sessions: light-to-moderate sessions of up to about 60 min did not appear to lower glycogen when intake was very high (between-study comparison; SYNTH-34 for the taper itself).
Women: eat the same per kilogram; the shortfall in older studies was the amount eaten, not the sex (REF-1205, REF-1206). At 8-10 g/kg/day two studies (McLay 2007 and REF-1206) found no evidence of a menstrual-phase difference (the review's narrative; SYNTH-37 coaches the cycle by symptoms and holds no glycogen data).
Very high intakes and the gut: the authors advise low-fibre, high-glycaemic sources to make 8-12 g/kg/day tolerable; rehearse the menu in a training week (practice, not tested here; SYNTH-02 for the race itself).
Only for long events: loading improved performance only past about 90 min (REF-819, B+); for shorter races normal high-carbohydrate eating is enough, and the review itself says normalising stores (6-10 g/kg/day) suffices for many sports.
Limitations
Heterogeneity. I² was 92-94% overall and stayed high in most subgroups; the pooled 191.9 is an average over very different protocols, not a predicted gain for one athlete.
Small, confounded subgroups. 2-10 comparisons per subgroup, mostly below the 10 the authors themselves cite as the minimum; duration is confounded by dose, and meta-regression was judged inappropriate.
Outcome. Vastus lateralis glycogen, not race or time-trial performance; the review does not test whether the extra glycogen changes performance.
Who. 88% men; no group older than 37; the well-trained regression rests on few study groups with intakes between 4.7 and 12.0 g/kg/day.
Real-world fit. 65% of trials rested participants during loading; diet was supplied in most trials but compliance was rarely measured.
Method. Values digitised from figures for 17 studies; non-independent observations in the regressions; no GRADE or publication-bias assessment (the authors' own limitation).
Funding and interests: the funding statement says the authors have nothing to report; no conflicts declared.
How this entry was checked and changed
Why this level
SR/MA (n = 641 participants in 63 trials; 18 trials pooled) with considerable heterogeneity (I² 92-94%), mostly small crossover trials and subgroups of 2-10 comparisons; read from the full text (open access, CC BY 4.0). The outcome is muscle glycogen, not performance, and no certainty-of-evidence grading (GRADE, publication bias) was done
Full text (checked 2026-10-26)
Descriptive synthesis included 63 trials (n = 641 participants), 18 were included within the meta-analysis (full text)
Raising relative intake from below to above 6.5 g/kg/day increased muscle glycogen by 191.9 mmol/kg DM (95% CI 142.8 to 241.1, n = 111) in repeated measures trials and by 107.6 (95% CI −10.5 to 225.7, n = 81, p = 0.07) in parallel group trials (full text)
Heterogeneity was I² = 94% for repeated measures and I² = 92% for parallel group designs (full text)
Loading duration subgroups: 24 h 207.7, 48 h 255.0, 72 h 148.0 and 96 h 106.4 mmol/kg DM; subgroup differences p = 0.01 (full text)
In well-trained participants relative intake predicted only 8.6% of the variation in muscle glycogen (p = 0.147) (full text)
Only 74 females were included across all trials (12% of participants) (full text)
Athletes should consume > 8 g/kg/day for 36-48 h pre-competition, with prior glycogen-depleting exercise and high glycaemic index carbohydrate allowing 24-36 h (full text)
The true loading period in Bussau et al. was ~36 h, as the final habitual training session was ~12 h before loading (full text)
Of the 42 randomized experiments 24 posed some concerns; of the 21 non-randomized trials 13 posed a serious risk of bias (full text)