REF-1364 · Jones et al., 2026 · International Journal of Sport Nutrition and Exercise Metabolism · Paper · PubMed

Does carb-loading still help if I fuel well during the race?

Unproven: of 14 loading trials only 2 were double-blind, and both, with carbohydrate during exercise, found no performance gain. Small evidence (B-, from the abstract).

Main caveat: Abstract only. The full text is behind the publisher's paywall, there is no Europe PMC copy, and the University of Southern Denmark portal record listed by…
limited, mixed evidenceMeasured on recreational to highly trained athletes

Jones, Louis, Saunders, Morton, Gejl, Nielsen and Areta start from the standard view that carbohydrate loading (CL) improves time to fatigue and gives modest performance gains in exercise lasting about 90-240 min, and ask whether that benefit could be a placebo effect, given that belief and expectation can change performance on their own.

They systematically evaluated loading research for two features: double-blind, placebo-controlled (DBPC) design, and carbohydrate provided during the exercise test. Of 14 studies identified, only two were DBPC trials and only four provided carbohydrate during exercise.

Both DBPC trials included appropriate carbohydrate intake during exercise and reported no evidence of a performance benefit of loading.

In the other two studies that fed carbohydrate during exercise, loading improved performance, with a moderate effect that the authors call comparable to that reported for placebo nutritional interventions (Cohen's d ≈ 0.35; the abstract does not make clear which of the two the figure describes). The studies also differed in test type (time to exhaustion vs time trial).

The authors conclude that whether loading in the 24-48 h before competition enhances performance, once placebo effects are accounted for and in-race fuelling is optimised, needs re-evaluation, and they call for more DBPC trials.

Our reading: this is the base's named dissent on the second half of PHYS-04 ("a fuller store helps only past about 90 min") — not a finding that loading does nothing, but an audit showing that the performance claim rests almost entirely on trials that were not double-blind, most without race fuelling, and that the two trials designed to remove expectation found no gain.

Two trials of unknown size cannot settle it either way.

Protocol

Design: commentary with a systematic evaluation of carbohydrate-loading studies; outcome of the evaluation = counts of design features (DBPC or not; carbohydrate during exercise or not; performance test type), with the reported performance effects of the subsets.
Sample: n = 14 studies (participant numbers not given in the abstract).
Comparison inside the audit: loading vs a control diet in each study (the control conditions are not described in the abstract); DBPC trials vs the rest; trials with vs without in-exercise carbohydrate.
Not given in the abstract: the search strategy and inclusion criteria; the identity, size, population, loading protocol and event of each study, including the two DBPC trials; how blinding of a high-carbohydrate diet was achieved; the effect sizes in the DBPC trials; funding and competing interests.

Key Results

Item (abstract)Count or result
Loading studies identified14
Double-blind, placebo-controlled2 of 14
Carbohydrate given during exercise4 of 14 (the two DBPC trials and two others)
Studies with neither feature10 (14 − 4; our arithmetic, since both DBPC trials fed carbohydrate during exercise)
The two DBPC trials (both with appropriate in-exercise carbohydrate)no evidence of a performance benefit of loading
The two other trials with in-exercise carbohydrate (not double-blind)a positive performance effect, moderate and comparable to placebo nutritional interventions (Cohen's d ≈ 0.35, which the abstract places after the placebo comparison)
Test types across studiesheterogeneous: time to exhaustion vs time trial

Mechanisms

Authors' interpretation: belief and expectation can independently influence performance, so in unblinded loading trials part or all of the reported benefit may be a placebo effect; carbohydrate eaten during the race should be optimised before loading is judged.
Measured (in the included trials, as summarised): the two trials that blinded the diet and fed carbohydrate during exercise found no performance benefit.
Inference: carbohydrate taken during exercise supplies fuel that may offset part of a lower starting glycogen store, which would shrink the gain from loading in a well-fuelled race; the abstract does not state this mechanism, and it was not tested here.
Inference: the physiology that loading raises muscle glycogen is not in dispute (REF-1204, a meta-analysis of 63 biopsy trials sharing its first author and Areta); what is disputed is how much extra performance a fuller store buys once expectation and race fuelling are controlled.

Practical Application

Does carb-loading still help if I fuel well during the race? Unproven: of 14 loading trials only 2 were double-blind, and both, with carbohydrate during exercise, found no performance gain. Small evidence (B-, from the abstract).
What it is good for as a source: naming the dissent on loading's performance benefit in its authors' terms: the benefit has been shown almost only in trials that were not double-blind, mostly without race fuelling. It does not show that loading is useless, and its two null trials are of unknown size.
How a principle should cite it: "the two double-blind loading trials, both with carbohydrate during exercise, found no gain (Jones 2026, from the abstract)" — not "loading does not work".
What it means for training: Not from this paper; the base's own practice: loading for events beyond about 90 min stays the default where it costs little (SYNTH-64, dose from REF-1204 and REF-748), and fuelling during the race is never replaced by it (SYNTH-02); expect a smaller gain than older trials that were not double-blind suggest.

Limitations

Abstract only. The full text is behind the publisher's paywall, there is no Europe PMC copy, and the University of Southern Denmark portal record listed by OpenAlex holds no file. The 14 studies, the two DBPC trials, their participants and loading protocols, the search method and the funding and competing interests could not be read.
A count of designs, not a meta-analysis. The paper's central point rests on two trials; no pooled effect, confidence interval or risk-of-bias rating is given in the abstract. No evidence of benefit in two trials is not evidence of no benefit (our inference).
The placebo comparison is an analogy. That d ≈ 0.35 is "comparable" to placebo nutrition effects is the authors' comparison with other literature, not a test inside these trials.
Framed as a commentary that calls for more DBPC trials and proposes research directions; the authors re-open the question rather than answer it.
Same group as [REF-1204]. Not a conflict of interest, but the two papers are not independent readings of the loading literature.
How this entry was checked and changed
Why this level

a structured audit of 14 loading studies by a leading carbohydrate group (Liverpool John Moores University and the University of Southern Denmark; its first author, Jones, and Areta are also authors of REF-1204). This is a commentary, not a narrative review, so the leading-group line (B+) does not apply; it is graded B- (abstract only, key methods missing, conflicting results) because the point the base cites it for — no performance gain in the two double-blind trials — rests on two trials whose size, population and protocol the abstract does not give, and because the paper counts designs rather than pooling outcomes. The full text was not read (paywalled; no open copy)