Effects of Acute Caffeine Supplementation on Physical, Sport-Specific, Physiological, Perceptual, and Cognitive Outcomes in Female Team-Sport Athletes: A Three-Level Meta-Analysis

What Hedges g is

A way of expressing a difference in standard deviations rather than in seconds or centimetres, so results from unlike tests can be pooled. Roughly, 0.2 counts as small and 0.5 as moderate.

Why they ran it

Caffeine has a substantial evidence base in sport but its applicability to female team sport athletes is insufficiently defined, and the authors cite an audit in which women were 23 percent of participants across 1826 supplement studies.

Earlier female focused reviews were limited by few female only trials and by two level models that treat multiple correlated effects from the same participants as independent, which overstates precision.

Drawn from the paper's introduction, which states an aim rather than a hypothesis.

The reviewers looked for randomised, placebo controlled crossover trials of a defined caffeine dose in female team sport athletes, and found 26. Basketball, volleyball, handball, soccer, softball, rugby and water polo were all represented, mostly at elite or well trained level. Doses ran from 1.92 to 9 mg per kg, usually as capsules, taken a median of 60 minutes before testing. Because each trial contributes several results at once, the analysis used a three level model that keeps effects from the same study grouped rather than treating them as independent.

Physical performance came out at a Hedges g of 0.32, with a confidence interval of 0.22 to 0.42. Inside that, repeated sprint ability scored 0.43, agility and change of direction 0.42, sprint speed 0.39, anaerobic power 0.30 and jumping 0.29. Perceptual responses improved by 0.42, driven mainly by lower ratings of perceived exertion at 0.35. Sport specific skills pooled at 0.36, but the interval ran from minus 0.01 to 0.73, so it did not clear the line. Heart rate rose on caffeine, which is what a stimulant does.

Two things temper this. Not one of the 26 trials reached low risk of bias, mostly for thin reporting of randomisation and of how results were selected, and three were rated high risk because caffeine is hard to hide from the person taking it. That matters most for perceived exertion and for all out efforts, where belief can move the number. The cognitive estimate rests on four studies, which the authors rate as very low certainty and ask readers not to treat as a benefit.

The numbers

undefined	undefined
undefined	undefined
undefined	undefined
undefined	undefined

Why this might happen

Proposed by the authors This is the explanation the authors offer in their discussion. This study did not test it.

The authors explain the physical results by caffeine acting as a non selective adenosine receptor antagonist that relieves central inhibition, increases motor cortex excitability and motor unit recruitment, and may improve calcium handling.

For the perceptual result they argue the same adenosine antagonism attenuates central fatigue signalling and the perception of effort, so a workload that feels easier can be sustained longer as fatigue accumulates.

They point to their own pooled data as consistent with that reading, since perceived exertion fell while heart rate rose slightly, though the receptor level steps were not measured.

Drawn from the discussion sections on physical performance and on perceptual responses.

What this does not show

This cannot support a dose or a timing window. Doses spanned 1.92 to 9 mg per kg and timing ranged from 15 to 70 minutes before testing, yet moderator analyses found no reliable moderation by dose, timing, formulation, sport, competitive level, habitual intake or blinding status.
It does not show caffeine sharpens thinking in these athletes. Only four of the 26 studies reported any cognitive outcome, with very wide uncertainty and very low certainty of evidence. The authors state directly that these data should not be read as showing a cognitive benefit.
It says nothing about menstrual cycle phase or hormonal contraceptive use. Reporting of both was too inconsistent across the included trials to support any subgroup or meta regression analysis, which is why the authors call for future trials to record them prospectively.
The perceived exertion finding is not a clean measurement. Caffeine produces noticeable alertness and tremor, so participants can often tell which arm they are in. The reviewers downgraded certainty for that domain specifically because blinding integrity may have been compromised.

Where this leaves us

It is now more likely that acute caffeine gives female team sport athletes a small, outcome specific benefit concentrated in explosive and repeated high intensity efforts and in lower perceived exertion, and not a general benefit for technical skills or cognition. It also contradicts any expectation of a tidy dose response.

Young adult and adult female team sport athletes, mostly elite or well trained, given a single acute dose and assessed largely in laboratory style tests rather than competitive match play. The authors say the findings should not be extended to adolescent athletes.

Adequately powered female specific trials that prospectively verify menstrual cycle phase, hormonal contraceptive use and habitual intake, prespecify sport specific skills and cognition as primary outcomes, and power genotype as an effect modifier rather than adding it afterwards.

Caveats worth holding

Two databases were searched, PubMed and Web of Science, so trials indexed only elsewhere or published in other languages may be missing from the pool.
Heterogeneity was high in several domains and the prediction intervals were wide, which means the next well run trial could land a long way from the pooled average.
Small study effects could not be ruled out. The funnel asymmetry tests came back clean but had little power, and the authors record publication bias as undetected rather than absent.
The individual trials were mostly too small for the job. The reviewers estimate their power to detect a small effect at around 8 to 9 percent, and no outcome reached 80 percent power.