REF-1683 · Tarrit et al., 2026 · The Journal of Sports Medicine and Physical Fitness, · Paper · PubMed

Carbohydrate during endurance exercise helped performance significantly in 102 studies (44% by the authors' count), with intake in 1-2 h and 2-4 h efforts appearing to favour performance and glucose-plus-fructose blends doing better than simple carbohydrate, in an umbrella review of 15 systematic reviews (8 with meta-analysis; 262 RCTs, mostly time trials) whose quality was low or critically low in 11 of 15 (AMSTAR2); the abstract counts significant trials and gives no pooled effect size

Main caveat: The authors': "more rigorous research" is needed to carry the results into athletes' real practice.
moderate evidenceMeasured on non-athletes to trained athletesRead from the abstract only

Tarrit and colleagues collected systematic reviews and meta-analyses of randomised controlled trials on carbohydrate intake during endurance exercise, to work out under which conditions it helps performance most. They searched five databases to September 2024 and included 15 reviews, 8 of them with a meta-analysis, which together covered 262 RCTs and 521 single trials.

They rated the quality of 11 reviews (73%) low or critically low with AMSTAR2. Performance was measured as a time trial in 69% of cases and as time to exhaustion in 31%. By their count, 102 studies (44%) showed a significant benefit of carbohydrate, with no difference between performance tests or sports.

A glucose-plus-fructose blend, which the abstract calls complex carbohydrate, improved performance compared with simple carbohydrate or other combinations, and intake during efforts of 1-2 h (57 trials) and 2-4 h (70 trials) appeared to favour better performance.

The authors conclude that carbohydrate is most useful for efforts of 1 to 4 hours and in blended formulations, and that "more rigorous research" is needed to bring the findings into athletes' real practice.

Participants

Pooled populations: endurance athletes, mainly cyclists or runners, across 262 RCTs and 521 single trials.
Number of participants, sex, age and training level: not in the abstract.

Protocol

Search: five databases (not named in the abstract), from inception to 5 September 2024, for systematic reviews and meta-analyses of RCTs on carbohydrate intake and performance.
Included: 15 reviews, 8 with meta-analysis.
Quality appraisal: AMSTAR2 for each review.
Outcomes: time-trial performance (69% of measurements) and time to exhaustion (31%).
Synthesis: counts of trials with a significant positive effect, by test type, sport, carbohydrate type and effort duration; no pooled effect size in the abstract.
Not in the abstract: the databases, doses and timing, the risk of bias of the trials, how overlap between reviews was handled, registration.

Key Results

FindingFigure in the abstract
Reviews included15 (8 with meta-analysis), 262 RCTs, 521 single trials
Review quality (AMSTAR2) low or critically low11 of 15 (73%)
Performance testtime trial 69%, time to exhaustion 31%
Studies with a significant positive effect of carbohydrate102 (44%, per the abstract; 102 of 262 RCTs is 39% by our arithmetic, so the denominator is unclear)
Difference by test type or sportnone
Glucose + fructose vs simple carbohydrate or other combinationsblend better (no effect size)
Effort duration where intake seemed to favour performance1-2 h (57 trials) and 2-4 h (70 trials)
Efforts shorter than 1 h or longer than 4 h: no result in the abstract.
The abstract gives no grams per hour and no effect sizes.

Practical Application

Take carbohydrate during efforts of about 1 to 4 hours; in the trials these reviews gathered, intake during 1-2 h (57 trials) and 2-4 h (70 trials) efforts appeared to favour performance, and blends of glucose and fructose did better than simple carbohydrate or other combinations. This rests on reviews mostly rated low or critically low in quality, and on counts of significant trials rather than a pooled effect, so it shows the direction, not the size.
How much per hour is not answered here. The abstract gives no doses; the base's rows by duration and level stay those of SYNTH-02 (consensus 30-60 g/h for 1-2.5 h, 90 g/h beyond 2.5 h, from REF-748).
For the 40-km to 50-mile time trial this agrees with [SYNTH-62], which fuels the longer efforts: 60 g/h from about 90 min, and 30-60 g/h or a mouth rinse below that. Inference, not tested here: this umbrella's 1-2 h band supports taking carbohydrate in an hour-plus time trial rather than relying on a rinse alone, though it gives no dose and no result for efforts under an hour.
Glucose plus fructose above about 60 g/h is already the base's rule (SYNTH-02, REF-748); this review's "complex carbohydrate" is that blend, not starch.

Limitations

The authors': "more rigorous research" is needed to carry the results into athletes' real practice.
Quality of the evidence: 11 of 15 included reviews rated low or critically low by AMSTAR2.
Synthesis: the abstract reports counts of significant trials (vote counting), which ignores effect sizes and penalises small trials; the denominator of the 44% is not stated (102 of 262 trials would be 39%), and the overlap of trials between the 15 reviews is not reported, so some trials may be counted more than once.
Not in the abstract: doses, timing, participants, sex and training level, the risk of bias of the trials, and results for efforts under 1 h or over 4 h.
Laboratory time trials and time-to-exhaustion tests, not races.
Written from the abstract and the PubMed record; full text not read.
How this entry was checked and changed
Why this level

an umbrella review, the rubric's A+ design, whose abstract reports its AMSTAR2 appraisal and the result (low or critically low in 11 of 15 reviews, 73%), so the abstract-only cap does not apply but the appraisal is serious and caps it at A-; held at B+ because the synthesis the abstract reports is a count of trials with a significant effect rather than a pooled estimate, the risk of bias of the underlying trials and the overlap of trials between reviews are not reported, and the abstract does not name the denominator of its '102 studies (44%)' (102 of 262 RCTs would be 39%, our arithmetic)