The following databases were searched: PubMed, CNKI, Web of Science, and EBSCO. A total of 32 articles were retrieved, with the search period spanning from the establishment of each database to 3 July 2024. The English search terms employed were as follows: (“blood flow restriction training” or “KAATSU training” or “BFR” or “pressure training”) and (“muscle strength” or “explosive force” or “muscle recruitment” or “explosive power” or “fatigue” or “tiredness” or “weariness” or “upper limbs” or “upper extremities”), and (“RCT”).

The retrieved literature was uniformly imported into EndNote software and screened by two researchers, J.W. and J.X., separately. Discrepancies were resolved through discussion and negotiation or third-party judgment. The literature screening and inclusion steps are shown in Figure 1. Finally, 32 articles were included.

Two researchers extracted data from the literature that met the inclusion criteria using a customized form, mainly including the following contents.

1. General information: first author, publication year.

Statistical analysis was performed using Revman 5.4. The standardized mean difference (SMD) and 95% confidence interval (95% CI) were selected as the effect size for the combined effect size. The quality of the included literature was evaluated using the Cochrane Risk of Bias Assessment Tool. Before the combined meta-analysis, a heterogeneity test was conducted first. The homogeneity test (Q test, test level a = 0.1) was used for the heterogeneity test. The value range of I ² was between 0% and 100%. When I ² > 50% and p < a, it indicated the presence of heterogeneity, and the random effects model was selected for the meta-analysis. Conversely, the fixed effects model was selected. The subgroup analysis was employed to address heterogeneity, and STATA 16.0 was utilized for sensitivity analysis to examine the stability of the results. The Egger test and funnel plot were adopted to verify the existence of publication bias.

A total of 32 publications were incorporated into this study. All of them were randomized controlled trials (RCTs), involving 524 subjects of mixed gender and with an age range of 16–45 years. The basic characteristics are presented in Table 1.

The quality of the literature was evaluated by referring to the Cochrane Risk of Bias Assessment Tool (Higgins and Altman, 2007). Seven aspects, including random sequence generation, allocation concealment, participant blinding, outcome blinding, incomplete outcome data, selective reporting, and other bias, were assessed using Review Manager 5.4 software (Figures 2A,B). Twenty-four articles failed to clearly describe whether the allocation personnel strictly adhered to random allocation, and 32 articles were at high risk of bias in blinding due to the signing of informed consent forms before the experiment.

Among the 32 included articles, 17 articles compared the upper limb muscle strength of 341 subjects before and after BFR training (Figure 3). Heterogeneity test showed that I ² = 77% > 50%, and P of Q test <0.01, which means there is strong heterogeneity among the literatures. Therefore, random effects were selected for meta-analysis. The pooled SMD value of 17 studies is 0.36, and the 95% confidence interval is from 0.02 to 0.70, and it is statistically significant (Z = 2.06, P = 0.04), indicating that BFR training can significantly enhance the muscle strength of the upper limbs.

Regarding the impact of BFR training on fatigue degree, a total of 25 articles and 307 subjects were included (Figure 4). The heterogeneity test result is I ² = 88% > 50%, and p of Q test <0.01, indicating significant heterogeneity among studies. Therefore, a random effects model is used for analysis. The results show that the combined effect size of BFR training on fatigue degree is SMD = 1.38, 95%CI (0.81, 1.94), which is significant (Z = 4.79, p < 0.0001), suggesting that BFR training can affect fatigue degree to a large extent.

Based on the data of the impact of BFR training on upper limb muscle strength, subgroup analyses were conducted on Exercise mode, Exercise intensity, and Compressive strength respectively according to the characteristics that may cause heterogeneity (Table 2).

Subgroup analysis of different Exercise modes shows that the within-group heterogeneity of the group with Exercise mode as Elbow flexors has increased (I ² = 79%), compared with the overall combined effect (I ² = 82%), suggesting strong heterogeneity among the literatures of this Exercise mode. The number of studies on Pull-up and Wrist curl is small and lacks representativeness. Among them, Bench press has the highest homogeneity and is statistically significant (I ² = 0%, p = 0.02 < 0.05), indicating that upper limb BFR training can significantly improve bench press strength.

Subgroup analysis of Exercise intensity shows that the heterogeneities of the three groups are 34%, 0%, and 49% respectively, which are all reduced compared with the overall combined effect (I ² = 77%). It is speculated that different Exercise intensities may be one of the influencing factors leading to heterogeneity among studies. Compared with ≤30% 1RM and body weight exercises, the exercise intensity of 40%–70% 1RM reaches the maximum effect size SMD = 1.16 and is significant (p < 0.01), indicating that BFR training combined with 30%–70% 1RM can significantly enhance muscle strength. On the contrary, when the exercise intensity reaches Maximal effort, the effect size is negative SMD = −0.54 and is statistically significant p = 0.002 < 0.05, suggesting that exhaustive BFR training may temporarily have a negative effect on upper limb muscle strength.

The Compressive strength subgroup shows that the between-group heterogeneity of the group with applied BFR pressure ≤40% AOP has increased (I ² = 86%), suggesting strong heterogeneity among the literatures under this applied BFR pressure. Among them, the combined effect size of the group with ≥60% AOP is the highest and is significant (SMD = 0.64, P = 0.002), indicating that BFR training with ≥60% AOP can significantly improve upper limb muscle strength.

For the research on fatigue degree, the author suspects that the sources of heterogeneity are the inconsistency of measurement tools, exercise intensity, and compressive strength. Therefore, subgroup analysis is carried out according to the source ofheterogeneity (Table 3).

According to different measurement tools, among the four groups of Rate of Perceived Exertion (RPE), muscle soreness, Visual Analogue Scale (VAS), and blood lactate, the heterogeneity between studies of the RPE and VAS scales is relatively high (90% and 98% respectively). Among them, the combined effect sizes of RPE (SMD = 1.11, P = 0.004) and blood lactate (SMD = 2.15, P < 0.0001) are statistically significant and positive, indicating that upper limb BFR training can significantly increase the RPE and blood lactate of the subjects.

In different Exercise intensity subgroups, the heterogeneities of the four groups are 91%, 0%, 0%, and 0% respectively. Compared with the overall combined effect (I ² = 88%), the 40%–70% 1RM, Maximal effort, and Self weight groups show a completely homogeneous state. It is speculated that different exercise intensities may be an influencing factor leading to heterogeneity among studies. When the exercise intensity is Maximal effort, the effect size is the largest and significant (SMD = 2.61, P < 0.0001). When the exercise intensity is Self weight, the effect size is negative and statistically significant (SMD = −1.14, p = 0.0002). The results show that upper limb BFR training combined with maximal effort exercise intensity will significantly increase the fatigue degree of the subjects, while upper limb BFR training combined with body weight exercise intensity is not easy to cause fatigue in the body.

The Compressive strength subgroup shows that the heterogeneities of the three groups are 88%, 75%, and 88% respectively. Among them, when the applied BFR pressure is ≥ 60% AOP, the effect size is the highest and significant (SMD = 2.64, P = 0.0001). This indicates that under different applied BFR pressures, there are differences in the impact on the fatigue degree of the subjects. A higher applied BFR pressure (≥60% AOP) can significantly increase the fatigue degree of the subjects, while the effects under other applied BFR pressures are relatively weaker.

Sensitivity analysis was conducted on the included studies, and individual studies in each group were excluded to determine heterogeneity.

Table 4 shows that the combined effect size of BFR training on upper limb muscle strength is [SMD = 0.36, 95% CI (0.02, 0.70), p = 0.04, I ² = 77%]. After excluding the studies of Jessee, Lambert, and Vivian, I ² decreased to 65%, 76%, and 70% respectively, and heterogeneity was reduced to a certain extent. After excluding individual studies, the range of combined effect size SMD is between 0.29 and 0.46, and the p-values are all less than 0.1, which fully shows that no single literature will pose a threat to the results of meta-analysis.

Table 5 indicates that after excluding individual studies on the impact of individual BFR on fatigue degree, the range of the standardized mean difference (SMD) is between 1.25 and 1.53, and the p-values are all less than 0.01. Compared with the combined effect size [SMD = 1.38, 95% CI(0.81, 1.94), p < 0.0001, I ² = 88%], the research results remain unchanged, suggesting that this study has good stability.

Offset tests were conducted separately according to the subgroups of the effects of BFR training on upper limb muscle strength and fatigue degree. The publication bias of the included literatures was investigated by drawing funnel plots (Figures 5, 6). The test results show that the funnel plots are basically symmetrical, and the p-values in the Egger test are all greater than 0.05. This fully indicates that within the scope of this study, there is no obvious publication bias in the studies on the effects of BFR training on upper limb muscle strength and fatigue degree, and the research results have high reliability.

Strength is critical for athletic performance (Young, 2006). This study examined acute effects of BFR training combined with various protocols on upper limb strength. Meta-analysis of 18 studies demonstrated significant positive effects (p < 0.05), indicating BFR training’s efficacy in enhancing muscular strength. The mechanism involves limb occlusion-induced distal congestion, creating localized hypoxia and lactic acid accumulation that recruits fast-twitch fibers (Yazgan et al., 2024), consistent with previous findings (AmiltonVieira et al., 2015). Notably, high-threshold motor unit activation depends on contraction dynamics and intramuscular oxygen levels (Lau et al., 2023; Toshiaki et al., 2016). BFR-induced metabolite accumulation may accelerate neuromuscular activation through: 1) metabolic stimulation of group III/IV afferents (Jnsson et al., 2024), and 2) cross-bridge cycle modulation (Kubota et al., 2008), highlighting its potential for strength training.

Sensitivity analysis confirmed result stability, though substantial heterogeneity (I ² = 82%) suggests methodological variations across studies. This variance may stem from multiple factors. Firstly, BFR training programs vary in applied pressure, intensity, duration, and frequency, all of which can impact training outcomes (Jennings et al., 2005). Secondly, different measurement methods, with their diverse levels of accuracy and reliability, also affect results (Batista Mauro et al., 2011). Thus, this study performed subgroup analysis based on potential sources of heterogeneity.

Muscle fatigue, a physiological phenomenon during exercise, is characterized by reduced muscle strength, exercise capacity, and subjective tiredness (Constantin, 2024). Our findings confirm that BFR training can significantly influence the degree of fatigue. This result can be understood from multiple perspectives. During exercise, reduced oxygen and nutrient supply to muscles shifts energy production to anaerobic pathways (Kof and Strojnik, 2008). Local hypoxia and metabolite accumulation alter nerve cell ion channel function, impairing nerve impulse conduction (Gamboa and Andrade, 2012). Some studies have pointed out that in BFR training, the firing frequency of motor neurons is reduced and the activation degree of muscles is decreased (Martin et al., 2015; Fahs et al., 2015), which may be one of the reasons for fatigue.

Although BFR training significantly impacts fatigue, the results show high heterogeneity, likely due to varied intervention methods, training intensities, durations, and applied pressures. Differences in method reliability and effectiveness, as well as higher compression levels causing severe local hypoxia and metabolic stress, may exacerbate fatigue (Godawa et al., 2012). To address this, subgroup analyses were conducted based on measurement tools, exercise intensities, and BFR pressure.

Despite rigorous screening and exclusion, several limitations remain. Ethical guidelines require participants to provide informed consent, disclosing study details and intervention specifics, which inherently compromises blinding. In this review, 33 studies could not implement blinding due to this requirement. While this methodological limitation exists, it is unlikely to severely distort results, given the emphasis on participant safety and transparency. Additionally, although BFR training has been shown to improve strength and power, the lack of standardized protocols leads to variability in training programs and testing methods, contributing to data heterogeneity. Future research should refine experimental designs to better understand the impact of factors like training status and exercise intervals.