Caffeine (1,3,7-trimethylxanthine) is one of the most widely consumed performance-enhancing substances in sport due to its well-established ergogenic effects. The use of caffeine is more common in aerobic-based sports due to the ample evidence endorsing the benefits of caffeine supplementation on endurance exercise. However, most of this evidence was established with cycling trials in the laboratory, while the effects of the acute intake of caffeine on endurance running performance have not been properly reviewed and meta-analyzed. The purpose of this study was to perform a systematic review and meta-analysis of the existing literature on the effects of caffeine intake on endurance running performance. A systematic review of published studies was performed in four different scientific databases (Medline, Scopus, Web of Science, and SportDiscus) up until 5 October 2022 (with no year restriction applied to the search strategy). The selected studies were crossover experimental trials in which the ingestion of caffeine was compared to a placebo situation in a single- or double-blind randomized manner. The effect of caffeine on endurance running was measured by time to exhaustion or time trials. We assessed the methodological quality of each study using Cochrane’s risk-of-bias (RoB 2) tool. A subsequent meta-analysis was performed using the random effects model to calculate the standardized mean difference (SMD) estimated by Hedges’ g and 95% confidence intervals (CI). Results: A total of 21 randomized controlled trials were included in the analysis, with caffeine doses ranging between 3 and 9 mg/kg. A total of 21 studies were included in the systematic review, with a total sample of 254 participants (220 men, 19 women and 15 participants with no information about gender; 167 were categorized as recreational and 87 were categorized as trained runners.). The overall methodological quality of studies was rated as unclear-to-low risk of bias. The meta-analysis revealed that the time to exhaustion in running tests was improved with caffeine (g = 0.392; 95% CI = 0.214 to 0.571; p < 0.001, magnitude = medium). Subgroup analysis revealed that caffeine was ergogenic for time to exhaustion trials in both recreational runners (g = 0.469; 95% CI = 0.185 to 0.754; p = 0.001, magnitude = medium) and trained runners (g = 0.344; 95% CI = 0.122 to 0.566; p = 0.002, magnitude = medium). The meta-analysis also showed that the time to complete endurance running time trials was reduced with caffeine in comparison to placebo (g = −0.101; 95% CI = −0.190 to −0.012, p = 0.026, magnitude = small). In summary, caffeine intake showed a meaningful ergogenic effect in increasing the time to exhaustion in running trials and improving performance in running time trials. Hence, caffeine may have utility as an ergogenic aid for endurance running events. More evidence is needed to establish the ergogenic effect of caffeine on endurance running in women or the best dose to maximize the ergogenic benefits of caffeine supplementation.

Flexibility training in swimming is a key to improving the flexibility index of athletes. In this era of rapid development of IoT technology, it is necessary to combine the Internet of Things technology with the application of wearable sensors in swimming flexibility training. This article is to study the value of wearable sensors in swimming flexibility training under the background of the Internet of Things. In this paper, by combining the interconnected nature of the Internet of Things technology, choosing Apache MINA and the XStream library platform to construct the Internet of Things system, combined with the data measurement characteristics of wearable sensors, an improved strategy is analyzed by measuring various indicators in the flexibility training of swimmers and through the movement interference removal combined with the adaptive algorithm to reduce the energy consumption in flexibility training. In this paper, two sets of flexibility training experiments of static stretching and dynamic stretching are designed. In the experiment, the flexibility index and energy consumption of athletes are received through the wearable sensor of the Internet of Things, and finally, the new improved strategy is analyzed through the received data. And this article also designed a set of control experiments. The experiment shows that wearing IoT wearable sensors can improve the flexibility index of swimmers up to 20.41, while the highest is 15.53 without wearing them.

Objective: To evaluate and compare the effectiveness of different physical training environments and modalities on swimming performance and sport-specific skills in competitive swimmers using a two-tier network meta-analysis.Methods: A systematic search of six databases identified 36 randomized controlled trials involving 844 competitive swimmers. A first-tier network meta-analysis compared aquatic, dry-land, and combined training environments across performance outcomes (25 m, 50 m, 100 m, 200 m times) and sport-specific metrics (start time, turn time, swim velocity, stroke rate, stroke length). A second-tier analysis further examined specific training modalities within combined and dry-land categories. Mean differences (MD) and standardized mean differences (SMD) with 95% confidence intervals (CI) were reported; interventions were ranked using surface under the cumulative ranking curve (SUCRA).Results: Combined training showed the highest efficacy across multiple outcomes. Compared to control, it significantly improved 100 m time (MD = −2.01 s; 95% CI: −2.87 to −1.16), swimming velocity (MD = 1.27 m/s; 95% CI: 0.61–1.94), stroke rate (SMD = 1.63; 95% CI: 0.92–2.34), and stroke length (SMD = 0.86; 95% CI: 0.23–1.49). In the second-tier analysis, water plus dry-land resistance training (W + DRT) ranked highest across 50 m, 100 m, swim velocity, and stroke metrics. Core training and power training showed specific benefits for 25 m sprint (MD = −0.90s; 95% CI: −1.79 to −0.01) and take-off velocity (MD = 0.18 m/s; 95% CI: 0.03–0.32).Conclusion: Combined aquatic and dry-land training—especially W + DRT—most effectively improves swimming performance and sport-specific skills. Core and power training function as targeted adjuncts. These findings provide a concise, precision-based prescription for physical preparation in competitive swimming.

Objective: To identify upper-trunk–pelvic roll coordination strategies during 400-m front-crawl swimming and examine their kinematic characteristics, segment-wise distribution, and associations with swimming speed. Methods: Twenty-two elite male swimmers completed a maximal 400-m front-crawl trial while two inertial measurement units recorded upper-trunk and pelvic roll. Cycle-level cross-correlation lag was used to classify coordination into upper-trunk-leading (UL), pelvic-leading (PL), and near-synchronous (NS) strategies. Kinematic differences among strategies and changes in strategy probability across 50-m segments and swimming speed were examined using Bayesian multilevel models. Results: Across 2,796 valid roll cycles, NS was the most frequent strategy (53.04%), followed by PL (25.86%) and UL (21.10%). Strategy use varied substantially between swimmers. UL showed greater torso twist range than NS (NS–UL = −0.71, 95% CrI: −0.96 to −0.42) and greater torso twist velocity than both NS (NS–UL = −0.61, 95% CrI: −0.95 to −0.26) and PL (UL–PL = 0.62, 95% CrI: 0.19–1.04). In contrast, upper-trunk and pelvic roll ranges were larger in NS and PL than in UL, with no clear difference between NS and PL. Faster within-athlete speed increased the odds of UL relative to both NS (OR = 2.38, 95% CrI: 1.76–3.26) and PL (OR = 2.07, 95% CrI: 1.31–3.33). No clear segment effects or between-athlete mean speed effects were observed. Conclusion: Elite 400-m front-crawl swimmers used three distinct upper-trunk–pelvic lead–lag strategies, with marked inter-individual variability. NS was most common, whereas UL was more likely at faster within-athlete speeds. These findings suggest that trunk coordination in elite swimmers is individualized rather than organized around a single optimal pattern.

The elite swimming calendar became substantially denser after 2020. We examined how the individual-event medalist population at the nine global championships of 2016–2025 changed in age structure, podium tenure, medal accumulation, and turnover. All 865 individual-event medals (277 athletes) at three Olympic Games and six World Championships were organized as an athlete-by-edition panel and analyzed with athlete-clustered and wild cluster bootstrap inference, exploratory Bayesian partial pooling, discrete-time hazard models, negative binomial regression, and bootstrap measures of concentration. Mean medalist age ranged from 22.86 to 24.46 years across editions. None of the 34 sex-by-event era contrasts survived false-discovery-rate adjustment; the two with the smallest p-values (men’s 200 m breaststroke, +2.79 years; women’s 400 m individual medley, −5.19 years; both q = 0.062) are exploratory, and both were accompanied by near-complete turnover of the event’s medalists. Median podium tenure was two editions. Each additional year of age at the first observed medal was associated with a higher hazard of a last observed medal (HR 1.106, 95% CI 1.063–1.151) and lower subsequent accumulation (IRR 0.863, 95% CI 0.823–0.904), whereas winning two or more medals at the first observed championship was associated with about half the hazard (HR 0.564, 95% CI 0.364–0.874) and twice the accumulation rate (IRR 2.066, 95% CI 1.410–3.027). The title-defense rate was 58.8% across 2017 → 2019 and 15.4% across the later World Championship transitions (OR 0.128, 95% CI 0.040–0.409), while the Gini coefficient of medal counts ranged from 0.205 to 0.248. Medal concentration remained descriptively similar, while the identities of the winners changed more frequently.

Objective: To identify data-driven Dmax-derived threshold-intensity profiles in adolescent swimmers by jointly considering pool-based and land-based incremental-test indicators and to examine hematological, performance-related, maturation-related, and psychoperceptual characteristics associated with profile membership. Methods: This multicenter cross-sectional study included 247 adolescent swimmers from three training centers. Participants completed a land-based cycling incremental test and a pool-based 5 × 200 m front-crawl incremental test. Pool-based Dmax-derived threshold intensity was expressed as swimming velocity at the Dmax point relative to event-specific personal-best velocity, whereas land-based Dmax-derived threshold power was expressed as cycling power output at the Dmax point. The two indicators were entered into latent profile analysis. Between-profile comparisons incorporated effect sizes, 95% confidence intervals, and Benjamini–Hochberg adjustment for multiple testing. Candidate variables were selected according to prior evidence and biological plausibility, screened for strong pairwise correlations, and entered into a multinomial logistic regression model with study center included as a categorical fixed effect. Results: Three ordered threshold-intensity profiles were identified: lower (n = 117, 47.4%), intermediate (n = 72, 29.2%), and higher (n = 58, 23.5%). The selected model showed good classification accuracy (entropy = 0.929). Relative to the lower profile, membership in the intermediate profile was associated with a higher red blood cell count (OR = 7.765, 95% CI: 3.198–18.849), lower perceived exertion (OR = 0.500, 95% CI: 0.372–0.672), and greater last-repetition completion relative to personal-best velocity (OR = 1.725, 95% CI: 1.317–2.259). Membership in the higher profile was associated with a higher red blood cell count (OR = 40.412, 95% CI: 10.844–150.605), lower perceived exertion (OR = 0.300, 95% CI: 0.197–0.457), greater last-repetition completion (OR = 2.903, 95% CI: 2.007–4.200), lower mental fatigue (OR = 0.954, 95% CI: 0.916–0.993), and greater sport mental toughness (OR = 1.134, 95% CI: 1.041–1.235). Conclusions: Adolescent swimmers exhibited three data-driven, ordered patterns in the joint distribution of pool-based Dmax intensity and land-based Dmax power. Profile membership was associated with concurrent hematological, performance-related, and psychoperceptual characteristics. These profiles provide an exploratory person-centered framework for describing cross-modal threshold-intensity heterogeneity rather than biologically discrete phenotypes or prescriptive training categories. Their reliability, longitudinal stability, and responsiveness to training require further evaluation.