REF-1741 · Salame, Brown, Oueijan & McCullough, 2026 · Journal of the International Society of Sports Nutrition, · Paper · PubMed
Carbohydrate during cycling in the heat helped in 5 of 9 small crossover trials and not in 4 (gains of 13.4-19.3% in time to exhaustion and 3.3-12.7% in time trials where significant; doses 14-140 g/h at 27-35 °C), with no pooled estimate; the 140 g/h arm tended to shorten time to exhaustion and only 2 trials asked about gut symptoms. Systematic review, read in full
Main caveat: Nine small crossover trials (6-16 participants); no meta-analysis, so the synthesis is a count of significant results and cannot say how large the average…
Salame and colleagues (Leeds Beckett University and the American University of Beirut) asked whether carbohydrate taken during endurance exercise improves performance in the heat, where faster glycogen use has been proposed as one reason performance falls.
From 3151 records they kept 9 randomised crossover trials, all cycling in climate chambers at 27-35 °C and 20-78% humidity, with average intakes from 14 to 140 g/h and exercise of about 50 to 152 min.
Five trials found a significant benefit (time to exhaustion 13.4-19.3% longer; time-trial performance 3.3-12.7% better) and four found none; the authors judged the heterogeneity in protocols, doses, conditions and outcomes too great to pool. Of eight trials with indirect calorimetry, five found no difference in carbohydrate oxidation.
Carbohydrate did not change hydration, core temperature or perceived exertion in most trials, and only two trials measured gut symptoms. The authors conclude that carbohydrate in the heat does not consistently improve performance, and point athletes to hydration, pre-exercise fuelling and gut training from the wider literature.
Participants
9 trials, 6-16 participants each; mostly men under 30; study mean VO2max about 52.4-66.0 mL/kg/min (one trial's mean age 45 ± 4 years).
Women: 2 in one series of one trial and 7 of 14 in another; the rest all-male.
Acclimatisation: four trials described their participants as non-acclimatised; five gave no data.
Inclusion: healthy, at least recreationally active, aged 18-65, non-smokers, not pregnant, without metabolic, gut, cardiovascular, neurological or psychological disease.
Protocol
Search: four databases to 14 April 2026, plus reference lists and grey literature; 3151 records, 1689 screened after duplicates, 51 sought in full (one unobtainable), 9 included.
Inclusion: carbohydrate during exercise against a control differing only in carbohydrate (combinations with protein or caffeine allowed); continuous endurance exercise over 30 min with a direct performance test (time trial or time to exhaustion); ambient temperature over 23 °C, real or simulated.
Interventions: 8 trials used drinks against a flavour-matched placebo or water, one used sago porridge against nothing; three tested a single sugar (glucose or maltodextrin, 49.5-99.0 g/h), five tested blends containing fructose or sucrose (14-140 g/h), one sago (62 g/h).
Tests: four time-to-exhaustion trials (60-73% VO2max, or 55% of peak power) and five steady rides of 45-120 min followed by a time trial of about 3 to 28 min.
Risk of bias: RoB 2 by two reviewers, disagreements settled by consensus or a third reviewer; extraction by one author, checked by a second.
Synthesis: narrative; differences computed as carbohydrate minus placebo, in absolute terms and as a percentage of placebo.
Funding and interests: no funding; no conflicts reported.
Key Results
Each trial's performance result as tabulated by the review:
| Trial | Conditions | Exercise and test | Intake (g/h) | Change vs control | Significant |
|---|---|---|---|---|---|
| Abbiss | 32 °C, 55% | 90 min at 62% VO2max + 16-km TT | ~75 sucrose | 3.3% faster (26.6 vs 27.5 min) | yes |
| Carter (a) | 35 °C, 30% | to exhaustion at 60% and 73% VO2max | ~58 maltodextrin | +18.3% and +19.3% | yes (p = 0.028, 0.04) |
| Carter (b) | 35 °C, 30% | to exhaustion at 55% peak power | ~58 maltodextrin, sweet or not | +18.8% and +13.4% vs water | yes |
| Che Jusoh | 30 °C, 78% | 45 min at 55% VO2max + 15-min TT | ~62 sago | −0.9% (219 vs 221 kJ) | no |
| Cureton | 28.5 °C, 60% | 120 min at 60-75% VO2max + 15-min TT | 63.5 sucrose/dextrose | +6.7% (190 vs 178 kJ) | no |
| Davis | ~27-28 °C, 67% | 1 h at 65% VO2max + ~3-min sprint | 99.0 or 49.5 glucose | −0.7% and +0.3% | no |
| Febbraio | 33 °C, 20-30% | to exhaustion at 70% VO2max | 140, 70 (series 1); 42, 70 (series 2) | −24.5%, −7.6%; +7.4%, +7.3% | no (140 g/h tended lower, p = 0.09) |
| Flood | 32 °C, 70% | 90 min at 45% VO2max + 15-min TT | 90 maltodextrin/fructose | +12.7% vs water (169 vs 150 kJ) | yes (p = 0.005) |
| Watson | 30 °C, 60% | to exhaustion at 60% VO2max | 14, 29, 43 sucrose/glucose/fructose | +10.2%, +11.6%, +18.5% | only 43 g/h (p = 0.045) |
Secondary outcomes:
Carbohydrate oxidation (8 trials): no difference in 5; higher in 3 (the 140 g/h drink, the 90 g/h blend, and a trend reaching significance at two time points with 58 g/h maltodextrin).
Hydration: fluid intake was similar between conditions; sago kept plasma volume better and the 140 g/h drink lost more plasma volume; the others found no difference.
Core temperature (8 trials): no difference in 7; higher with carbohydrate in one (Abbiss); peaks about 38-39 °C.
Perceived exertion (7 trials): no difference in 6; a trend to lower RPE in one.
Gut symptoms (2 trials): 99 g/h glucose gave more nausea, fullness and stomach upset than 49.5 g/h and placebo; the 90 g/h blend gave more fullness, with no significant difference in gut disturbance, against water.
Risk of bias: overall "some concerns" in all trials; randomisation and outcome measurement raised concerns in about 45% of trials each (read from the figure), period and carryover, deviations and missing data low. The text puts every trial's overall rating down to one item of the selective-reporting domain (no analysis plan finalised before unblinded data), while the figure shows that domain as low risk in all trials.
Practical Application
In the heat, carbohydrate helped in about half of the small laboratory trials and harmed in none significantly; the significant gains came at about 43-90 g/h, though several null results fell in the same range, and the one arm at 140 g/h tended to shorten time to exhaustion. No pooled effect; all cycling, mostly drinks, mostly non-acclimatised men.
[SYNTH-02] hot-day row agrees in direction: keep the row's rate if rehearsed in heat and do not raise it. The 140 g/h result (p = 0.09) and the extra gut symptoms at 99 g/h of glucose alone both point against raising the rate in heat.
Where this review differs from [SYNTH-02]: the guide's hot-day instruction is an inference from one glucose-only heat-versus-cool trial (REF-461); it cites this review's nine trials of carbohydrate against a control at 27-35 °C (14-140 g/h, five of them sucrose or fructose blends) in its hot-day row and "In the heat", and lists a glucose-fructose blend at 100-120 g/h in heat as untested, with 90 g/h tested once against plain water.Among them, one trial of a 90 g/h maltodextrin-fructose drink at 32 °C and 70% humidity raised 15-min time-trial work by 12.7% against plain water (not a taste-matched placebo), with more fullness but no significant gut disturbance, and one trial of 140 g/h of glucose, sucrose and glucose polymer tended to shorten time to exhaustion against 70 g/h and a sweet placebo (p = 0.09); each is a single small trial.
Where the authors' advice differs from [SYNTH-02]: they put hydration first and question routine carbohydrate supplementation in the heat because benefits were inconsistent; SYNTH-02 keeps the duration-and-level rate. Inference, not tested here: a count of five positive and four null trials, none significantly harmful, gives no effect size; it reads as weak support for fuelling at moderate rates rather than as evidence against it.
Inference, not tested here: rehearse the hot-day rate in the heat; most trials did not ask about gut symptoms, so the tolerable rate in heat for a given rider is unknown.
Limitations
Nine small crossover trials (6-16 participants); no meta-analysis, so the synthesis is a count of significant results and cannot say how large the average effect is.
Mixed tests: four time-to-exhaustion trials, which the authors note repeat poorly, and five time trials; three of the four time-to-exhaustion trials and two of the five time trials were positive, and the authors say the test type does not by itself explain the split.
All cycling in climate chambers, seven with fans; acclimatisation unreported in five trials; two trials of about 1 h may have been too short to stress glycogen, and one gave a large breakfast.
Eight of nine interventions were drinks; gels and solid food are barely covered.
Gut symptoms measured in 2 of 9 trials.
Internal inconsistencies in the paper: the abstract names "the Cochrane Collaboration's tool" and the methods RoB 2; the risk-of-bias section speaks of "ten eligible studies" against nine included; the text gives one water-controlled trial while the table lists water as the control in two; the discussion gives a 32% fall with the 140 g/h drink where the table gives 24.5%; the Davis trial is dated 1988 in one place and 1998 in another.The per-trial figures above follow the review's Table 2.
The review's supplementary risk-of-bias table was not read; the domain split is read from the figure description.
How this entry was checked and changed
Why this level
a systematic review whose risk of bias we read in the full text (every trial "some concerns" overall, which the authors attribute to one item on a prespecified analysis plan; none high risk), held at A- because the trials are all small crossovers and the synthesis is a count of significant results with no pooled effect