This systematic review and meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines (Supplementary Table 1) (39) and was pre-registered in PROSPERO (CRD420251025291). A systematic search was conducted in PubMed, Web of Science, Embase, Scopus, the Cochrane Library, ProQuest, and EBSCO from inception to May 1, 2025. Search strategies were adapted for each database using Boolean operators and wildcards. The detailed search strategies for each database are provided in Supplementary Table 2. All retrieved records were exported to a CSV file for deduplication in EndNote X9 (Clarivate Analytics, New York, NY, United States). Study identification, screening, and final selection were independently completed by two reviewers (ZZK and YMY), with disagreements resolved through discussion.

The inclusion and exclusion criteria were defined according to the PICOS framework (Supplementary Table 3) as follows: (1) Participants were adult female basketball players (18 years of age or over) across different training and competitive levels. In studies including both sexes, only data from female participants were extracted; when unavailable, corresponding authors were contacted for data acquisition (4, 24, 30), studies were excluded if the required data remained unobtainable after contact (40). Studies involving para-athletes were also excluded; (2) Interventions examined the acute effects of caffeine. Longitudinal studies assessing chronic caffeine intake, trials involving multi-ingredient supplements in which caffeine was not the primary active component, and studies lacking a clearly documented caffeine supplementation protocol were excluded; (3) Studies that used a design including a placebo condition as a comparator to caffeine; (4) crossover studies that compared the intake of caffeine and a placebo; (5) Studies using a blinded and randomized design; (6) Studies reporting at least one measure of athletic performance (e.g., vertical jump, physiological indicators). Unpublished data, grey literature, systematic reviews, and meta-analyses were excluded.

Data extraction was independently conducted by two authors (ZZK and QBP) using a customized Excel form developed prior to full-text screening. Information regarding menstrual cycle phase and oral contraceptive use was extracted when reported. For studies with missing data or outcomes reported exclusively in graphical format, corresponding authors were contacted via email to obtain the required information. If no response was received and the necessary data were available in figures, numerical values were extracted using WebPlotDigitizer 4.1. All digitized data were independently verified by two researchers, with any discrepancies resolved through consensus.

In this systematic review, for studies reporting post-intervention group means and standard deviations (caffeine vs. placebo), these values were directly extracted to calculate the standardized mean difference (SMD). When outcomes were reported only as pre–post measurements, baseline and final measurements were extracted to derive the mean change using the following equation (41):

where M diff denotes the mean change, M post the reported post-intervention mean, and M pre the reported pre-intervention mean.

The standard deviation of the mean change ( S D diff ) was then calculated as follows (41):

Where r represents the pre–post correlation coefficient. In accordance with the Cochrane Handbook, a conservative value of 0.5 was adopted in this study (41). The calculated mean change and its standard deviation were subsequently used to compute the standardized mean difference (SMD) for the meta-analysis.

The methodological quality of the included studies was assessed using the PEDro scale. Based on total scores, studies were classified as excellent (9–10), good (6–8), fair (4–5), or poor (<4). The PEDro scale has demonstrated good reliability and validity for evaluating the internal validity of randomized controlled trials (42).

Risk of bias was assessed using the RoB 2 tool in accordance with Cochrane guidelines (43). The evaluation encompassed the following domains: (1) bias arising from the randomization process; (2) bias due to deviations from intended interventions; (3) bias due to missing outcome data; (4) bias in outcome measurement; and (5) bias in selection of the reported results. For crossover trials, period and carryover effects were additionally examined as potential sources of bias. Each study was rated as having low risk, some concerns, or high risk of bias.

Both the PEDro and RoB2 tools were applied by two independent researchers (ZZK and MFH), with any disagreement resolved through consensus (Supplementary Table 4).

The certainty of evidence was assessed using the GRADE approach and categorized as high, moderate, low, or very low (44). The evaluation followed structured criteria: (1) Risk of bias: The certainty was downgraded by one level for outcomes rated as having “some concerns,” and by two levels for those rated as “high risk.” (2) Inconsistency: The certainty was downgraded by one level when heterogeneity fell between 25 and 75% (I² = 25–75%), and by two levels when heterogeneity exceeded 75%. (3) Imprecision: The certainty was downgraded by one level when the results lacked statistical significance. (4) Publication bias: In accordance with the Cochrane Hand-book (45), formal assessments of publication bias (e.g., Egger’s test) were not conducted due to the limited number of included studies (<10). Consequently, the certainty of evidence was not downgraded for publication bias. However, we acknowledge that the small number of included studies limits the statistical power to detect potential publication bias, which should be considered a limitation of this review rather than evidence of its absence (Supplementary Table 5).

Given the limited and inconsistent control of menstrual cycle phase and the lack of reporting on oral contraceptive use across studies, these factors were prespecified as sources of physiological heterogeneity and were considered during qualitative data synthesis and interpretation. This limitation also informed the cautious evaluation of the certainty of evidence.

Data synthesis was conducted using the inverse-variance method under a random-effects model, implemented with the Der Simonian–Laird estimator (42). Between-study variance components (tau² and tau) and their confidence intervals were estimated using the Jackson method (44). Pooled effect sizes were calculated using standardized mean differences (SMD), and their corresponding 95% confidence intervals (95% CI). Given the generally small sample sizes across the included studies, the bias-corrected Hedges’ g was used as the primary effect size metric (hereafter denoted as ES), with interpretive thresholds defined as follows: <0.2 trivial, 0.2–0.5 small, 0.5–0.8 moderate, and >0.8 large (46).

Heterogeneity was assessed using the I² statistic, tau², and tau, and a prediction interval (PI) was calculated to estimate the potential range of true effects in comparable future studies (47). I² values were interpreted as 0–25% low, 25–75% moderate, and >75% high heterogeneity (48). Publication bias was evaluated via funnel plots and Egger’s test (49, 50) with p > 0.05 indicating no significant bias.

All statistical analyses and graphical outputs were performed in RStudio (v4.5.0) using the meta and metafor packages. Statistical significance was set at p < 0.05, while p-values between 0.05 and 0.10 were considered indicative of a potential trend. Because standardized mean differences are sensitive to very small within-group variability, particularly in single-study subgroups (k = 1), results from such subgroups were interpreted with caution and used for descriptive purposes only.

Leave-one-out sensitivity analyses were conducted for outcomes including three or more studies by iteratively removing one study at a time and re-running the meta-analysis to assess the influence of individual studies on the pooled effect estimates. For outcomes supported by only two studies, formal sensitivity analyses were not feasible. Therefore, the robustness of these findings should be interpreted with caution (51). Prespecified sensitivity analyses, defined a priori based on the RoB 2 risk-of-bias assessment, were conducted by excluding studies with concerns in Domain 5 (bias in selection of reported results) to assess whether pooled effect estimates were driven by these studies.

A systematic search of six electronic databases, including PubMed (n = 982), Web of Science (n = 708), EMBASE (n = 236), Scopus (n = 260), ProQuest (n = 153), and EBSCO-host (n = 111) and the Cochrane Library (n = 90), yielded a total of 2,540 records. One additional study was identified through manual searching of reference lists (52). After removing duplicates and screening for eligibility, seven studies met the inclusion criteria (2, 4, 38, 52–55) (Figure 1), comprising 101 female basketball players.

As summarized in Table 1, the general characteristics of the included studies cover: (1) Author (s), year, and country; (2) study design; (3) Participants’ characteristics (sample size and, where reported, training level, age, body weight, height, menstrual cycle phase); (4) Caffeine consumption or restrictions; (5) Intervention (form, timing, and dosage); (6) Comparator (s) (form, dosage, and protocol); (7) Study period and washout; (8) Performance and physiological outcomes; (9) Funding source.

The PEDro scores of the included studies ranged from 7 to 10. Three studies were rated as excellent, and four were rated as good (Supplementary Table 4.1). Four studies did not meet the requirement for assessor blinding (4, 38, 54, 55), and three did not clearly report whether allocation concealment was implemented (38, 54, 55). Two studies failed to ensure therapist blinding (38, 54). Additionally, one study did not achieve completeness of primary outcome assessment (52), and another did not include dropout data in the final analysis (38).

According to the RoB 2 assessment (Supplementary Table 4.2), three studies were judged to have a low risk of bias and three to have some concerns, while one study was rated as having a high risk of bias. Four studies exhibited some concerns in Domain 5 (bias in selection of the reported result) (4, 38, 54, 55). In addition, one study showed some concerns in Domain 1 (bias arising from the randomization process) (55), one study in Domain 2 (bias due to deviations from intended interventions) (54), and one study in Domain 3 (bias due to missing outcome data) (38), respectively.

Based on the PEDro and RoB 2 evaluations, a risk of bias plot was generated using the Robvis tool (56) (Figure 2). Sensitivity analyses excluding studies with concerns in RoB 2 Domain 5 are presented in (Supplementary Table 9). These concerns, particularly regarding selective reporting of results, were considered in the subsequent certainty of evidence assessment using the GRADE approach.

The seven included studies collectively involved 101 female basketball players, with individual sample sizes ranging from 9 (52) to 26 (55). Participants were between 18 and 24 years of age. Except for two studies in which the training level was not specified (38, 55), the remaining studies reported that athletes had more than four years of systematic training experience. Except for one study that neither restricted nor reported habitual caffeine intake (52), all studies included athletes classified as low-to-moderate daily caffeine consumers.

Geographically, two studies were conducted in Australia, with one study conducted in each of Poland, Serbia, China, Iran, and Spain. A summary of study characteristics is presented in Table 1.

The administered caffeine dose across the included studies ranged from 2.1 to 9 mg/kg. Six studies (2, 4, 38, 53–55) provided caffeine in capsule form, while one study (52) used caffeinated gum. Caffeine was administered 70 min before testing in two studies (38, 55), 60 min in four studies (2, 4, 53, 54), and 15 min in one study (52). All studies implemented a one-week washout period between experimental conditions. Regarding caffeine withdrawal, although specific procedures varied, most studies required participants to abstain from all dietary sources of caffeine for 48 h before testing. Detailed supplementation protocols and timing information are summarized in Table 1.

Among the included studies, four did not report the menstrual cycle phase or its use in scheduling the experimental sessions (38, 52–54). Two studies conducted assessments during the luteal phase (2, 4), although neither specified the method used to determine menstrual cycle phase. One study performed testing on the 10th day following the onset of menstruation using a calendar-based approach (55). The specific menstrual cycle phase and its application in each study are summarized in Table 1. Furthermore, none of the seven included studies explicitly reported or controlled for oral contraceptive use.

This review categorized a priori into physical performance and physiological outcomes based on their primary construct and measurement characteristics. Physical performance outcomes reflected externally observable, task-based measures of physical capacity (e.g., sprint speed, jump height, and power output), whereas physiological outcomes represented internally regulated perceptual, physiological, or biochemical responses (e.g., fatigue perception and physiological or biochemical markers). Caffeine-related adverse effects were also summarized. The forest plot illustrating the overall meta-analytic results is presented in Table 2, while the independent forest plots for each outcome measure are provided in Supplementary Figure 7. Following recommendations from previous studies, caffeine doses for the subgroup analyses were classified as ≤3 mg/kg (low), >3–6 mg/kg (moderate), and >6–≥9 mg/kg (high) (57–59). The results of the subgroup analyses are available in Supplementary Table 8.

Subgroup analyses were conducted on an exploratory basis to examine potential sources of heterogeneity (e.g., caffeine dose). Consistent with the Cochrane Handbook for Systematic Reviews of Interventions, these analyses should be interpreted with caution given the limited number of included studies and statistical power, and were not used to override the conclusions of the primary meta-analyses (60–62).

Leave-one-out sensitivity analyses (Supplementary Table 6) were conducted for outcomes including three or more studies. For jump, agility, and sprint performance, the direction of the pooled effects remained consistent after sequential removal of individual studies, indicating that the results were not driven by any single study. For fatigue-related outcomes, exclusion of Stojanovic et al. (2) resulted in a slight change in effect direction. Regarding anaerobic performance, omission of Mahdav et al. (38) led to a more pronounced change in effect direction, whereas exclusion of Stojanovic et al. (2) had only a minor impact. Additionally, for jump performance, the pooled effect became statistically significant after exclusion of Nieto-Acevedo et al. (53), for all other outcomes, statistical significance remained unchanged (p > 0.05).

In Sections 3.6.1–3.6.3, the combined sample size refers to the total number of participant-observations contributing to each meta-analysis outcome. When a single study reported multiple eligible outcomes or testing conditions within the same domain, these were included as separate effect sizes for the purposes of synthesis. Accordingly, the number of effect sizes may exceed the number of studies, and combined sample sizes may differ across outcomes.

This systematic review synthesized evidence from seven randomized crossover trials involving a total of 101 individual female basketball players across all included studies. It aimed to evaluate the acute effects of caffeine on sport-specific skills, physical performance, and physiological and biochemical outcomes, and to examine the moderating roles of variables such as caffeine dose and test duration. The meta-analysis indicated that no statistically significant differences were observed across any of the analyzed performance variables.

In addition, the overall low certainty of evidence was partly driven by concerns related to selective reporting bias. According to the RoB 2 assessment, several included studies (4, 38, 54, 55) were judged as having some concerns in Domain 5 (bias in selection of the reported result), indicating insufficient transparency regarding pre-specified outcomes or analysis plans. This raises the possibility that favorable or statistically significant findings were preferentially reported, while null or unfavorable results may have been omitted. Such selective reporting can lead to an overestimation of intervention effects in meta-analyses. In line with GRADE guidance, this limitation contributed to downgrading the certainty of evidence for risk of bias, and the pooled estimates should therefore be interpreted with caution.

In the present review, the meta-analysis did not demonstrate a statistically significant effect of caffeine on shooting accuracy among female basketball players. While Liu et al. (63) reported improvements in shooting accuracy following caffeine ingestion, several other studies have failed to observe meaningful effects of caffeine on accuracy-related outcomes (14, 24, 28, 30, 64, 65). These inconsistencies may be attributable to several confounding factors, such as baseline shooting proficiency or testing conditions, which are difficult to standardize across studies. Although caffeine has been suggested to influence alertness, cognitive processing, and hand–eye coordination during high-intensity exercise (66–68), the relevance of these mechanisms for shooting accuracy outcomes remains unclear. In addition, estrogen-mediated physiological differences lead to slower caffeine metabolism in women, potentially prolonging its beneficial effects on neuromuscular coordination and attenuating fatigue-induced declines in accuracy (69, 70). Such mechanisms may confer a physiological advantage for female athletes when performing precision-dependent tasks. Comparable caffeine-induced improvements in accuracy have been observed in other sports, including rugby passing, volleyball attacking, and serving accuracy in female table tennis players (19, 71, 72). However, most existing studies have not systematically examined sex-specific differences in shooting accuracy, and it remains unclear whether caffeine elicits distinct performance effects in male versus female athletes. Nevertheless, accuracy is a complex performance skill shaped by the interplay of multiple factors, including postural control, motor patterns, sensory feedback, technical proficiency, psychological regulation, and resistance to fatigue (73–75) which may further obscure the manifestation of caffeine-induced enhancements.

The present meta-analysis did not identify a statistically significant effect of caffeine on dribbling sprint speed in female basketball players, a finding consistent with studies involving mixed-sex and female team-sport athletes (28, 64). Dribbling is a complex basketball skill requiring multidimensional coordination, including direction-al changes, acceleration and deceleration, and continuous adaptation to dynamic environments (76). However, most existing studies have assessed dribbling using linear sprint tests, which have limited ecological validity compared with real game situations. Future research should incorporate more dynamic and game relevant assessment protocols, such as change of direction tasks, combined start stop actions, and tests integrating technical skill transitions, to more accurately evaluate the actual effects of caffeine on dribbling performance.

Both dribbling and shooting are precision-dependent basketball skills rooted in the technical proficiency acquired through long-term training. Consequently, caffeine is unlikely to directly enhance these sport-specific technical abilities. However, when athletes already possess well-developed technical and tactical foundations, appropriate caffeine supplementation may provide a meaningful marginal benefit to overall performance.

The meta-analysis in this review showed that the effect of caffeine on jump height in female basketball players did not reach statistical significance. Although the confidence interval included zero, indicating that the current evidence is insufficient to determine the true effect of caffeine on jump height in female basketball players, the point estimate of the pooled effect size demonstrated a moderate positive magnitude. This positive direction is consistent with the direction of effects reported in previous systematic reviews predominantly involving male or mixed-sex samples (24, 28, 73). The extremely large effect size observed for the Abalakov jump in the subgroup analysis should be interpreted with caution, as it is based on a single study with exceptionally low variability and may be associated with task- and measurement-specific characteristics, such as the inclusion of an approach phase and greater reliance on stretch–shortening cycle utilization. The absolute improvement was modest (approximately 2.3 cm), indicating that the large, standardized effect primarily reflects statistical scaling rather than a markedly enhanced physiological response to caffeine. Therefore, it should not be inferred that caffeine provides a greater ergogenic benefit for complex jumping tasks compared with simpler jump movements.

It is worth noting that some studies have shown that women exhibit a greater proportional area of type I muscle fibers in certain skeletal muscles compared with men (77, 78), and that type I muscle fibers demonstrate higher sensitivity to caffeine under in vitro conditions (79). Compared with male athletes, this muscle fiber characteristic may theoretically cause female basketball players to display different physiological responses to caffeine and may even confer a potential physiological advantage in basketball performance, where explosive jumping actions (e.g., rebounding) are critical. In addition, previous studies (26, 80, 81) have indicated that vertical jump performance typically reaches its lowest level in the early morning (most commonly between 08:00 and 12:00) and peaks in the late afternoon or early evening (most commonly between 16:00 and 20:00). Furthermore, three studies reported that a caffeine dose of 3 mg/kg significantly improved vertical jump height in both male and female basketball players in the morning (primarily around 09:00) (26), while also producing small-to-moderate improvements in the evening (19:00–21:00) (2, 24). This effect may be related to caffeine’s ability to enhance motor unit recruitment and increase muscle activation levels (22, 82), thereby improving muscle contractile properties and attenuating diurnal fluctuations in neuromuscular performance. Based on the above considerations, the authors speculate that caffeine ingestion in female basketball players may help maximize training adaptations across different times of day and provide marginal performance benefits during evening competition periods.

In our study, no statistically significant effect of caffeine on agility performance was observed in female basketball players. However, subgroup analyses suggested that caffeine dosage may represent an important moderating factor influencing agility performance. Interestingly, findings from the study by Quan et al. (54) suggest that the effect of caffeine on agility performance in female basketball players may follow an inverted U-shaped dose–response relationship. Specifically, during prolonged exercise, performance following the ingestion of 6 mg·kg⁻¹ caffeine was superior to that observed with both lower (3 mg·kg⁻¹) and higher (9 mg/kg) doses. This pattern may be partly attributable to the relatively limited ergogenic effects of lower doses (e.g., 3 mg/kg), as well as to an increased incidence of adverse effects at higher doses (e.g., 9 mg/kg), which could attenuate the beneficial influence of caffeine on agility-related performance (83). Notably, such a dose-dependent profile aligns with previous findings (57, 84) reported in female or mixed-sex athletic populations across other sport disciplines, in which moderate caffeine doses (e.g., 6 mg/kg) have been shown to confer greater improvements in agility or related performance outcomes, although such evidence has not yet been established in female basketball players. Meanwhile, agility performance itself exhibits a clear circadian rhythm, being typically lower in the morning (most commonly between 07:00 and 9:30) and reaching peak values in the late afternoon to evening (most commonly between 17:00 and 20:00) (80, 85, 86). Further research by Bougrine et al. (87) demonstrated that caffeine ingestion of 3 to 6 mg/kg at 08:00 effectively enhanced agility performance in female athletes, with a greater improvement observed at 6 mg·kg⁻¹, which is consistent with an inverted U-shaped dose–response relationship, whereas caffeine intake during the evening (2, 24, 26) resulted in minimal performance benefits.

This meta-analysis did not identify a statistically significant effect of caffeine on maximal sprint performance in female basketball players. However, one study (52) reported that a caffeine dose of 2.3–3 mg/kg produced a moderate improvement in sprint performance in female basketball players, suggesting that very low doses of caffeine (<3 mg/kg) may have meaningful ergogenic potential. Two studies (2, 26) examining the effects of a low caffeine dose (3 mg/kg) demonstrated significant improvements in 10 m (ES = −0.63; −4.1%) and 20 m (ES = −0.41; −2.8%) sprint performance in female basketball players, with these effects being independent of circadian rhythm variations; no comparable improvements were observed in male basketball players at this dos. In contrast, male basketball players required a higher dose of 6 mg/kg to elicit a significant improvement in 20 m sprint performance (25). These findings suggest a potential sex-specific effect of caffeine on off-ball sprint performance, whereby female basketball players may obtain meaningful benefits at lower doses, whereas male players appear to require higher doses to achieve comparable ergogenic effects. However, it should be clearly noted that whether this effect is influenced by training status remains uncertain, and thus these conclusions should be interpreted with caution.

The present meta-analysis showed that caffeine ingestion may exert a moderate and statistically significant effect on power output in female basketball players (p < 0.05). Subgroup analyses indicated that caffeine dosage may moderate this effect (between-subgroup difference p < 0.01), with moderate doses (>3–6 mg/kg) appearing to elicit greater improvements. In contrast, outcome measures, testing modalities, and training status did not significantly moderate the observed effects (all between-subgroup differences p > 0.05), although significant pooled effects were observed across multiple outcome measures. Notably, Quan et al. (54) reported that a moderate caffeine dose (6 mg/kg) resulted in significantly greater improvements in power output compared with both a lower dose (3 mg/kg) and a higher dose (9 mg/kg) in female basketball players, a finding consistent with previous studies (88, 89) in female team-sport athletes and female soccer players. In contrast, Mahdavi et al. (38) reported no significant effect of a moderate caffeine dose (5 mg/kg) on power output in female basketball players, which aligns with earlier studies (90, 91) showing no significant improvement in power output during the Wingate anaerobic test following caffeine ingestion at approximately 6 mg/kg. These discrepant findings may be closely related to differences in power assessment methods. Studies reporting positive effects of caffeine on power output, such as those by Quan et al., predominantly employed field-based performance tests (e.g., repeated sprint tests), in which power variables are typically derived from sprint performance. In contrast, studies reporting null effects, such as that by Mah et al., primarily used cycle ergometer–based assessments (e.g., the Wingate test) to directly measure mechanical power output. Taken together, the effects of caffeine on power output in female basketball players remain inconclusive and appear to depend on both caffeine dose and the method used to assess power performance. Future studies using standardized power assessment protocols are warranted to clarify the dose–response relationship between caffeine supplementation and power output in this population.

In the present meta-analysis, no statistically significant ergogenic effect of caffeine on anaerobic performance was observed in female basketball players. In addition, the change in effect direction observed in the sensitivity analysis likely reflects the limited number of remaining effect sizes and the influence of individual studies, rather than a definitive reversal of the effect. Subgroup analyses indicated that, based on the current evidence, the predominant energy system required by the testing task may represent an important moderating factor influencing the effects of caffeine on anaerobic performance. It should be noted that Mahdavi et al. (38) reported that caffeine ingestion at a dose of 5 mg/kg significantly reduced the change in blood lactate concentration (SMD _diff = −2.25) as well as the magnitude of lactate accumulation (SMD _diff = −0.53). In contrast, the study by Quan et al. (54) demonstrated that caffeine ingestion at different doses (3, 6, and 9 mg/kg) exerted no significant effects on overall blood lactate concentrations, nor on the magnitude or direction of post-exercise lactate responses in female basketball players. These discrepancies may be attributable to methodological differences between studies, including variations in exercise protocols (Wingate test vs. high-intensity intermittent exercise) and the timing of blood lactate sampling (5 min post-exercise vs. immediately after exercise). Moreover, the ecological validity of using cycling-based power tests to assess basketball-specific performance, which involves more complex physiological and neuromuscular demands during training and competition, remains questionable.

The results of the meta-analysis indicated that caffeine ingestion did not produce a statistically significant effect on perceived fatigue (RPE) in female basketball players. Among the studies included in the present analysis, four investigations (2, 38, 52, 54) consistently reported that, across a dosage range of 2.1–9 mg/kg, caffeine exerted no significant effect or only a trivial effect on RPE in female basketball players. These findings are largely consistent with the study by Bougrine et al. (57) which demonstrated that caffeine ingestion at doses of 3 mg/kg and 6 mg/kg did not significantly affect RPE in female team-sport athletes, regardless of whether it was administered in the morning or in the evening. However, it should be noted that Stojanović et al. (26) reported that 3 mg/kg of caffeine significantly reduced RPE in male basketball players when ingested in the morning (approximately 10:00 h), a finding that contrasts with the results observed in female athletes. This discrepancy suggests that the time-of-day effects of caffeine on RPE may differ between sexes, and that evidence supporting a fatigue-attenuating effect of caffeine in female basketball players remains inconsistent and inconclusive. Although the present findings indicate that caffeine ingestion does not exert a significant overall effect on perceived fatigue in female basketball players, its potential underlying mechanisms remain worthy of consideration. Previous research (18) suggests that caffeine may attenuate fatigue and enhance muscular endurance during high-intensity exercise through mechanisms such as antagonism of adenosine receptors, increased central nervous system excitability, and modulation of neuromuscular regulation, at least under certain conditions. However, fatigue is a multifactorial and task-dependent phenomenon, and the physiological demands imposed by exercise duration and intensity may diminish or even obscure these potential ergogenic effects. Furthermore, given the inherently subjective nature of fatigue-related perceptual measures, the ecological validity of such indicators in basketball-specific performance research should be interpreted with caution.

Similarly, the present study did not identify any statistically significant effects of caffeine intake on the physiological or biochemical indices of female basketball players. Notably, previous studies (92–94) have suggested that caffeine ingestion, typically within the range of 3 to 7 mg/kg, may enhance the efficiency of energy substrate utilization, such as by increasing fat oxidation and reducing glycogen utilization. Additionally, caffeine may partially optimize physiological regulatory processes related to exercise performance, including elevations in plasma free fatty acids and epinephrine levels. In conjunction with the findings from the two studies (38, 55) included in the present analysis, namely, that ingestion of 5 mg/kg caffeine promoted post-exercise lactate clearance in female basketball players without exacerbating oxidative stress following high-intensity exercise, these results suggest that caffeine may confer beneficial and safe physiological advantages for female basketball players. However, it should be clearly noted that, given the absence of statistically significant effects of caffeine on physiological or biochemical indices observed in the present study, and considering that direct evidence in female basketball players remains limited, whether these potential physiological effects can be translated into meaningful improvements in exercise performance in this population still requires confirmation through future high-quality and sport-specific research.

In this review, the primary findings discussed above are synthesized in the conceptual framework shown in Figure 3.

In the present review, one study (52) employed an ultra-low dose of caffeine (2.3 ± 0.2 mg/kg) administered in the form of caffeinated chewing gum to examine its effects on competitive performance in female basketball players. The results indicated small-to-moderate effect size improvements in sprint performance, agility, jump height, and free-throw accuracy; however, none of these changes reached statistical significance. Similarly, another study (95) using 3 mg/kg of caffeine chewing gum in male basketball players reported moderate effect size improvements in agility and jump performance, yet without statistically significant differences. Although the ergogenic dose of caffeine is commonly defined as 3–6 mg/kg (83), accumulating evidence (4, 73, 96, 97) suggests that very low doses (<3 mg/kg) may still elicit positive effects in certain sporting contexts. Taken together with findings from previous literature, the consistent trends toward performance enhancement observed under small sample size conditions, despite the absence of statistical significance, suggest that ultra-low-dose caffeine supplementation may exert potential ergogenic effects in female basketball players and possibly other athletic populations. These findings warrant further investigation using larger sample sizes and more robust study designs. Furthermore, all studies included in this review employed caffeine doses below 9 mg/kg. The prevailing view holds that very high doses of caffeine (>9 mg/kg) do not confer additional performance benefits (94), largely due to an increased likelihood of adverse effects. Nevertheless, the specific effects of such supra-high doses on female basketball players or team-sport athletes remain unclear and require further investigation.

Regarding ingestion timing and form, six studies included in this review administered caffeine in capsule form 60 min prior to exercise to coincide with the typical plasma caffeine peak window (30–60 min) (98). In contrast, the study employing caffeinated chewing gum administered caffeine 15 min before exercise, which is consistent with its pharmacokinetic characteristics, as caffeine delivered via chewing gum is rapidly absorbed through the oral mucosa, reaching peak plasma concentrations within approximately 15 min (99–101). Although this rapid absorption may shorten the required pre-exercise ingestion window, current evidence does not support a clear performance advantage of caffeinated chewing gum over traditional caffeine ingestion methods.

Notably, genetic polymorphisms related to caffeine may represent a potential mechanism underlying inter-individual variability in responses to caffeine supplementation (102–105), particularly involving genes associated with caffeine metabolism (e.g., CYP1A2) and adenosine receptor-mediated signaling pathways (e.g., ADORA2A). Although the relevance of these genetic factors has not been extensively studied in female basketball players, this area warrants further investigation. Future research should incorporate genetic analyses to determine whether specific genotypes are associated with differential ergogenic responses to caffeine in female basketball players.

Numerous studies investigating caffeine use in female athletes have suggested that training status and the menstrual cycle may modulate the ergogenic effects of caffeine (4, 24, 106–109). Among the studies included in the present review, only a subset controlled for the menstrual cycle (2, 4, 55), whereas participants were predominantly semi-professional and elite athletes. This methodological heterogeneity may, to some extent, account for the observed variability across study findings.

Available evidence indicates that athletes with higher training status tend to perform closer to the physiological and neuromuscular ceiling (110, 111), such that the ergogenic effects of caffeine often manifest as small effect sizes or marked inter individual variability (4, 24, 52, 106). In this context, if the menstrual cycle is not controlled, phase dependent hormonal influences on neuromuscular function (112) and energy substrate utilization (113) may further mask or attenuate the potential effects of caffeine. In contrast, among female athletes with lower training status, caffeine is more likely to elicit moderate or pronounced ergogenic effects, and these outcomes may be relatively less sensitive to menstrual cycle related fluctuations.

It is noteworthy that the menstrual cycle primarily consists of the follicular and luteal phases, which are characterized by fluctuations in estrogen and progesterone levels and may substantially influence physiological responses to exercise in female athletes (114, 115). Previous evidence (116–119) suggests that during the luteal phase, female athletes exhibit increased high-intensity running frequency, sprint count, and high-intensity running distance, along with enhanced aerobic capacity, whereas these performance indices show a statistically significant decline during the follicular phase. In contrast, shooting and rebounding performance appear to be superior during the follicular phase (120). Consequently, the ergogenic effects of caffeine supplementation may vary across menstrual cycle phases, and potential benefits could be attenuated or obscured by phase-dependent performance fluctuations.

From a mechanistic perspective, caffeine has been shown to increase intracellular Ca²⁺ and K⁺ availability. Enhanced Ca²⁺ release from the sarcoplasmic reticulum facilitates cross-bridge formation and contractile force production, whereas elevated K⁺ levels may stimulate Na⁺/K⁺-ATPase activity, thereby improving membrane excitability and delaying fatigue (121–123). In addition, estrogen and progesterone have been reported to exert opposite effects on muscle strength, with estrogen being positively associated and progesterone negatively associated with force production (119, 124, 125). Accordingly, female athletes in the follicular phase, characterized by elevated estrogen and relatively low progesterone concentrations, may exhibit greater strength and power, potentially amplifying the ergogenic effects of caffeine (126). Conversely, during the luteal phase, increased progesterone levels may attenuate neuromuscular performance and partially counteract or mask caffeine-induced benefits. Although some studies have reported no significant effect of menstrual cycle phase on strength performance, these hormonally mediated interactions remain a plausible regulatory mechanism and warrant further investigation.

Overall, training status may influence the magnitude of caffeine’s ergogenic effects, whereas menstrual cycle phase may affect their stability and direction. Therefore, without concurrent control or stratified analyses of these two factors, the true effects of caffeine on basketball performance in female athletes may be underestimated or appear inconsistent. This should be carefully considered when interpreting discrepancies across existing studies.

Common potential adverse events include tachycardia, headache, insomnia, and increased energy/activity levels, which may be related to the hormonal regulation of caffeine metabolism. The luteal phase is characterized by elevated estrogen and progesterone levels (127), and a reduced caffeine clearance rate (128). Under this high estrogen environment, sustained adrenergic stimulation may increase susceptibility to caffeine-related adverse events (129). Furthermore, it is worth noting that Nieto-Acevedo et al. (53) reported a higher incidence of increased urinary output in the placebo group compared to the caffeine group (20% vs. 0%, p < 0.05), whereas Stojanovic et al. (2) did not observe significant changes. This finding contrasts with those in male or mixed-gender cohorts (26, 64), which may be related to exercise intensity: high-intensity exercise promotes the secretion of antidiuretic hormone (ADH, i.e., arginine vasopressin, AVP) (130), and estrogen lowers the threshold for AVP release (131).

Overall, the effects of caffeine on female basketball players are influenced by multiple factors. To enhance performance while minimizing adverse reactions, it is advisable to personalize caffeine dosage, with a recommendation to test caffeine use during training before official competitions to assess individual responses and tolerance.

In the leave-one-out sensitivity analyses (Supplementary Table 6), changes in effect direction were primarily observed for anaerobic power and fatigue-perception performance. Such directional shifts are more likely to reflect evidence fragility and task- or measurement-related heterogeneity rather than statistical calculation errors. Given the limited number of studies contributing to these outcomes (only three for anaerobic power and four for fatigue-perception performance), and the fact that the pooled effects in the primary analyses were close to null, even minor differences in individual study effects were sufficient to shift the pooled estimates across the null value, indicating limited stability of meta-analytic results based on a small evidence base.

Moreover, the primary analysis of anaerobic power demonstrated substantial heterogeneity (I² = 86.6%), suggesting marked differences in how anaerobic capacity was defined and assessed across studies. For example, anaerobic outcomes derived from direct power output, composite performance tests, or metabolic response indices may not reflect the same underlying physiological construct. The subgroup analyses further support this interpretation, indicating that construct-level differences among anaerobic power measures amplified the influence of individual studies on the overall estimates. In contrast, fatigue-perception outcomes are inherently more subjective and therefore more susceptible to scale-related variability and methodological factors; the return of pooled effects toward null values after omitting individual studies further suggests that the evidence base for this outcome remains limited.

Overall, these directional changes indicate that, under conditions of a small number of included studies and heterogeneous outcome constructs, the available evidence for these outcomes remains fragile, and the corresponding conclusions should be interpreted with caution.

This systematic review has several limitations. First, only seven studies were eligible for inclusion, which limits the robustness of the pooled estimates and warrants caution when interpreting the findings in this specific population. While the focus on female basketball players is important, the current evidence base remains limited. Moreover, participant age and competitive level varied across studies, and although these characteristics were recorded when reported, potential age or expertise related differences could not be systematically examined and should be considered when interpreting the findings. Therefore, the present analysis should be regarded as a preliminary exploration of the potential performance effects of caffeine in female basketball players, and additional well-designed studies are needed to better understand and confirm these results. Second, substantial heterogeneity was observed in the performance assessment protocols, and some tests may have limited relevance to basketball-specific movements, potentially affecting ecological validity. Finally, basketball performance is shaped by interacting technical, physical, and tactical factors; thus, caffeine-related physiological effects represent only one component of performance and the present findings should not be overinterpreted.

Based on the available evidence, several avenues warrant further investigation: (1) the effects of caffeine on decision-making and physical function during critical phases of competition (e.g., end-game periods) or under acute fatigue; (2) interactions between hormonal fluctuations across the menstrual cycle and caffeine metabolism in female athletes, as well as the feasibility of adjusting caffeine intake strategies accordingly to sustain performance; (3) sex-specific mechanisms underlying caffeine responsive-ness, particularly how physiological differences shape individual response patterns; and (4) the potential impact of interactions between caffeine and genetic variability in female basketball players. These remain important yet underexplored areas for future research.