REF-1688 · Martins et al., 2026 · Nutrients, · Paper · PubMed

Both low (3 mg/kg or less) and moderate (4-6 mg/kg) caffeine doses shortened aerobic time-trial completion time against placebo (SMD −0.27, 95% CI −0.44 to −0.11, and SMD −0.52, −0.77 to −0.28), in a meta-analysis of 48 randomised placebo-controlled trials with 689 healthy adults; no eligible trial tested more than 6 mg/kg, and moderate doses helped trained and highly trained people alike. Risk-of-bias findings are not in the abstract

Main caveat: From the abstract: no study tested more than 6 mg/kg with a time-to-completion outcome, so the dose range above that is untested here.
moderate evidenceMeasured on non-athletes to highly trained athletesRead from the abstract only

Martins and colleagues asked whether higher caffeine doses add anything to time-trial performance, since larger doses may bring more side effects.

They pooled randomised placebo-controlled trials in healthy adults aged 18 to 59 who took oral anhydrous caffeine before an aerobic time trial, counting only time-to-completion outcomes, and grouped doses as low (3 mg/kg or less), moderate (4-6 mg/kg) and high (over 6 mg/kg). Forty-eight studies with 689 participants qualified.

Low doses shortened completion time against placebo (SMD −0.27, 95% CI −0.44 to −0.11, p = 0.001), and moderate doses with a larger point estimate (SMD −0.52, −0.77 to −0.28, p < 0.0001), the difference between doses not tested in the abstract. No eligible trial tested a high dose with a time-to-completion outcome.

With moderate doses, the effect was similar in trained and highly trained individuals. The authors conclude that low doses of 1.3-3 mg/kg may already improve endurance time trials, and that moderate doses of 4-6 mg/kg appear to give a more consistent benefit.

Participants

48 studies, 689 healthy adults aged 18-59.
Subgroups named: trained and highly trained individuals (definitions and numbers not in the abstract).
Sex split, exercise mode and distances: not in the abstract.

Protocol

Search: PubMed, Embase and Virtual Health Library; dates not in the abstract.
Inclusion: randomised placebo-controlled trials; healthy adults 18-59; acute oral anhydrous caffeine before an aerobic time trial; outcomes measured only as time to completion.
Dose groups: low ≤3 mg/kg, moderate 4-6 mg/kg, high >6 mg/kg.
Analysis: standardised mean differences with 95% CIs, random effects; subgroups by training status within the moderate dose.
Risk of bias: Cochrane Risk of Bias tool; findings not in the abstract.
Not in the abstract: crossover or parallel designs, timing of intake, habitual caffeine use, heterogeneity, publication bias, the low-dose trials' exact doses beyond the 1.3-3 mg/kg range in the conclusion, and registration.

Key Results

Dose vs placeboSMD for completion time (95% CI)p
Low (≤3 mg/kg)−0.27 (−0.44 to −0.11)0.001
Moderate (4-6 mg/kg)−0.52 (−0.77 to −0.28)< 0.0001
High (>6 mg/kg)no eligible studies—
Negative values mean faster completion with caffeine.
Moderate doses in trained and in highly trained individuals: similar effects (no figures in the abstract).
The abstract does not report a test of the difference between the low and moderate doses; the two intervals overlap (−0.44 to −0.11 and −0.77 to −0.28).
Background in the abstract: individual time-trial responses to caffeine range from about −3% to +16%.

Practical Application

Caffeine before a time trial helps at both low and moderate doses: 3 mg/kg or less shortened completion time (SMD −0.27) and 4-6 mg/kg somewhat more (SMD −0.52) across 48 placebo-controlled trials, with nothing tested above 6 mg/kg. The difference between the two doses was not tested in the abstract, so "moderate is better" is a direction, not a demonstrated gain.
This fits [SYNTH-67]'s 3-6 mg/kg default (REF-350: possibly effective from 2 mg/kg, about 9 mg/kg adds side effects, not benefit). Inference, not tested here: start a training trial at about 3 mg/kg and move toward 4-6 mg/kg if it is tolerated and the low dose did little; there is no evidence here for going above 6 mg/kg.
Individual responses vary widely (about −3% to +16% in time trials, the authors' background figure), so the group effect does not promise a gain to any one athlete; SYNTH-67's training trial decides, and one trial in the base found some men slower at 4 mg/kg by genotype (REF-910).
Trained and highly trained people both benefited from moderate doses, so training status is no reason to skip caffeine.

Limitations

From the abstract: no study tested more than 6 mg/kg with a time-to-completion outcome, so the dose range above that is untested here; the authors' own limitations are not stated.
Not in the abstract: risk-of-bias findings, heterogeneity, publication bias, the trial designs, exercise modes and distances, sex split, timing of intake, habitual caffeine use and registration.
The low-versus-moderate comparison is between groups of trials, not within trials, and no test of the difference is reported.
Trials overlap with other caffeine reviews in the base (likely, not reported).
Written from the abstract and the PubMed record; full text not read.
How this entry was checked and changed
Why this level

a meta-analysis of randomised placebo-controlled trials, the rubric's A design, capped because the abstract names the risk-of-bias tool but not what it found: risk of bias not read (abstract only); the trials are small (689 people across 48, about 14 per study by our arithmetic) and heterogeneity is not given