REF-1687 · Bittencourt de Santa Ritta et al., 2026 · International Journal of Sport Nutrition and Exercise Metabolism, · Paper · PubMed

Caffeine gave women a small, significant improvement in cycling performance (SMD 0.21, 95% CI 0.11 to 0.29; high certainty) in a three-level meta-analysis of 19 randomised crossover trials with 235 women; the effect did not differ significantly by dose or training level, and by cycle phase it was significant in the mid-luteal phase (SMD 0.26) and not in the others, which the authors say inadequate cycle assessment leaves unsettled; overall risk of bias "some concerns"

Main caveat: The authors': few clinical trials and inadequate menstrual cycle assessments preclude definitive conclusions about cycle phase.
fairly strong evidenceMeasured on non-athletes to trained athletesRead from the abstract only

Bittencourt de Santa Ritta and colleagues set out to fill the gap left by caffeine research done mostly in men, pooling randomised crossover trials that tested caffeine against placebo on cycling performance in females.

They searched three databases and included 19 trials with 235 women, analysing them with a three-level model so that several performance outcomes from one study could be kept.

Caffeine gave a small, significant ergogenic effect (SMD 0.21, 95% CI 0.11 to 0.29), which they rated high certainty, with no heterogeneity (Q(36) = 7.99, p = 1.00); the authors rated the overall risk of bias "some concerns". The effect did not differ significantly by caffeine dose (p = .79) or training level (p = .97).

By menstrual cycle phase, caffeine improved performance significantly in the mid-luteal phase (SMD 0.26, 0.06 to 0.46, p = .01) and showed no effect in the other phases (p > .11).

The authors caution that the few trials and inadequate menstrual cycle assessments rule out firm conclusions about phase, and conclude that caffeine can be used to improve cycling performance in women.

Participants

19 randomised crossover trials, 235 females.
Age, training level, sport background, contraceptive use and how the cycle phase was determined: not in the abstract.

Protocol

Search: three databases (not named in the abstract); studies selected by the PICOS framework and PRISMA guidelines; search dates and registration not in the abstract.
Analysis: three-level meta-analysis (performance outcomes nested within studies); standardised mean differences; subgroups by menstrual cycle phase, caffeine dose and training level.
Risk of bias and heterogeneity: RoB 2; Q test and funnel plot.
Not in the abstract: doses in mg/kg, timing and form of caffeine, the cycling tests, the phases compared and their verification, the number of trials per subgroup, and the certainty method (the abstract says "high certainty of evidence" without naming GRADE).

Key Results

AnalysisResult
Pooled effect of caffeine on cycling performance (19 trials, 235 women)SMD 0.21 (95% CI 0.11 to 0.29), significant; high certainty
Dose subgroupsno significant difference (p = .79)
Training-level subgroupsno significant difference (p = .97)
Mid-luteal phaseSMD 0.26 (0.06 to 0.46), p = .01
Other menstrual cycle phasesno effect (p > .11)
HeterogeneityQ(36) = 7.99, p = 1.00
Overall risk of bias (RoB 2)"some concerns"
Q of 7.99 on 36 degrees of freedom is far below what sampling error alone would give, which can arise from how the dependence of outcomes within studies was handled (our reading); the abstract does not discuss it.
The phase result is a subgroup comparison; the abstract does not give the number of trials per phase or the effect in the other phases.

Practical Application

Caffeine works for women cyclists too: across 19 crossover trials with 235 women it improved cycling performance by a small margin (SMD 0.21) against placebo, rated high certainty, with no significant difference by dose or training level. The default in SYNTH-67 (caffeine as the race supplement after a training trial) applies to women on this evidence rather than by transfer from men.
No reason to time caffeine by cycle phase. The mid-luteal benefit (SMD 0.26) and the null in other phases come from a subgroup split that the authors themselves say inadequate cycle assessments make inconclusive; this sits with SYNTH-37's rule that the phase is not the lever.
Dose: the abstract gives no mg/kg values, only that the dose subgroups did not differ (p = .79); the base's 3-6 mg/kg, possibly from 2 mg/kg (REF-350), stays the guide's range, and a lower dose tried first in training is the cautious start (inference, not tested here).
Gaps [SYNTH-67] names stay partly open: it lists women as thin for most supplements; this review covers caffeine in cycling only, and the base's genotype evidence (REF-910, REF-909) is still from men.

Limitations

The authors': few clinical trials and inadequate menstrual cycle assessments preclude definitive conclusions about cycle phase.
From the abstract: small trials (about 12 women per trial on average, our arithmetic); overall risk of bias "some concerns"; subgroup results are comparisons between groups of trials.
Not in the abstract: databases and search dates, registration, doses, timing and form of caffeine, the cycling tests, participants' age, training level and contraceptive use, how phases were verified, trials per subgroup, and publication-bias findings beyond the funnel plot being used.
The very low Q statistic (7.99 on 36 degrees of freedom) is unexplained in the abstract.
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 trials whose abstract reports its risk-of-bias result (RoB 2, overall "some concerns") and a high certainty rating, so the abstract-only cap does not apply; held at A- because the trials are small (235 women across 19, about 12 per trial by our arithmetic) and the phase and dose findings are subgroup splits