Low-Load Blood-Flow Restriction Exercise to Failure and Nonfailure and Myoelectric Activity: A Meta-Analysis

This systematic review was developed using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses. The review protocol was prospectively submitted to the PROSPERO platform (ID: 150824).

The electronic database searches included MEDLINE, PubMed, CINAHL, Web of Science, CENTRAL, Scopus, SPORTDiscus, PEDro, and the trial register in the ClinicalTrials.gov website (ie, gray literature). A systematic literature search was carried out with no language restriction between database inception (April 1978) and September 2019. The search terms were derived from blood flow restriction, blood flow restriction exercise, blood flow restriction training, kaatsu, vascular occlusion exercise, muscle fatigue, electromyography, emg, muscle activation, and clinical trial (see Supplementary Table, available online at http://dx.doi.org/10.4085/1062-6050-0603.20.S1). We also manually checked the reference lists of the identified studies, previous systematic reviews, and forward citations (to August 2020) for potentially relevant studies.

All studies were screened and assessed for eligibility according to inclusion and exclusion criteria based on the PICOTS principle (ie, extracting population, intervention, comparison intervention, outcome measures, time point measure, and study design information). Those involving participants (healthy or unhealthy, active or sedentary) aged ≥18 years were included. We considered interventions that compared LL-BFR (≤50% of 1RM/maximal voluntary contraction [MVC]) with LL-RT (≤50% of 1RM/MVC) or HL-RT (≥70% of 1RM/MVC).^1,2  The LL-BFR protocol needed to follow the same resistance-exercise modality (ie, isometric or dynamic) used in the LL-RT and HL-RT protocols, and LL-BFR and LL-RT needed to follow a matched number of repetitions (in nonfailure protocols) and load. Similar protocols have been deemed essential for drawing conclusions about the additional effects of BFR and enabling comparability. Surface EMG variables measured in time or frequency domains during MVC or exercise bouts were considered outcome measures. Only randomized crossover trials, randomized within-participant trials, or randomized controlled trials assessing short-term or acute responses, long-term adaptations, or both were included in this review. Studies involving a single exercise section for each intervention group were considered short term, and those involving at least 4 weeks of intervention were considered long term.

Those studies without available full text (short version) or without both LL-BFR and LL-RT conditions were not included. Studies involving only intermittent BFR or aerobic exercise as the intervention group, assessing only individual motor-unit characteristics with no EMG evaluation, or lacking the EMG signal in at least 1 agonist muscle of the movement performed were excluded.

The searches, data collection, risk-of-bias assessment, and data extraction were performed separately by 2 independent reviewers (D.G.M., J.A.M.B.), and when a consensus was not reached, the differences were discussed with a third evaluator (M.S.C.). The reviewers initially judged the relevance of the studies by reading the titles and abstracts. Those that did not match the inclusion criteria and duplicates were excluded. Articles with abstracts that presented the potential for eligibility or raised questions were retained for careful full-text analysis.

From the remaining eligible papers, the following data were extracted: (1) study design; (2) participant characteristics (sample size, sex, age, and training status); (3) exercise protocol (number of series and repetitions, frequency of training [number of days per week], training length [number of weeks], exercised muscle, muscle action mode [isometric, isokinetic, concentric, or eccentric], and exercise load); (4) cuff settings (pressure and width); (5) outcome measures (surface EMG variables extracted from time or frequency domain [assessed muscles and normalization method]); and (6) main study results. Data were extracted using a custom spreadsheet built by the third evaluator (M.S.C.).

Given the limited and heterogeneous data regarding frequency-domain measures, long-term studies comparing LL-BFR with HL-RT and studies assessing EMG time-domain variables immediately pre-exercise and postexercise bouts were analyzed only qualitatively. Thus, for the meta-analysis, the following comparisons were performed: (1) short-term LL-BFR versus LL-RT, (2) short-term LL-BFR versus HL-RT, and (3) long-term LL-BFR versus LL-RT effects. For these comparisons, time-domain variables (ie, muscle excitation) were extracted during submaximal exercise bouts and MVC in short- and long-term studies, respectively. For each comparison, subgroup analyses were conducted for both failure and nonfailure protocols.

The quality assessment was conducted using the Cochrane risk-of-bias tool, which classifies the risk of bias as high, low, or unclear. The risk of bias is considered high if a methodologic procedure is not described, unclear if the description is unclear, or low if the entire procedure is described in detail. Information from the studies was independently extracted by both reviewers and stored in Review Manager (RevMan) software (version 5.3; The Nordic Cochrane Center, The Cochrane Collaboration) for subsequent data crossing and discussion of possible discrepancies. Potential biases were assessed via visual inspection of funnel plots (Supplementary Figures 1 and 2).

We used the random-effects model in the meta-analysis to evaluate standardized mean differences (SMDs) ± 95% CIs. Those SMD outcomes between ≥0.2 and <0.5 were considered small, those between ≥0.5 and <0.8 were considered medium, and those ≥0.8 were considered large.^22,23  A random-effects meta-analysis was conducted because of data heterogeneity from different EMG variables.²⁴  Mean, SD, and sample-size data were extracted. Regarding the meta-analyses performed for studies assessing short-term responses at multiple time points, only the last time point was considered, because the variables were normalized for MVC or the beginning of the exercise. In studies assessing the long-term effects, the mean and SD difference values between pretraining (SD_pre) and posttraining (SD_post) were used. Given that we partially observed differences between time points (ie, between SD_pre and SD_post), we defined the SD_change as follows: square root ([SD_pre²/N_pre] + [SD_post²/N_post]), where N was sample size.²⁵  We either contacted authors when essential data were not comprehensively reported in the manuscript or estimated data from graphs using the ImageJ software (National Institutes of Health). If any information regarding SDs was missing, these values were calculated from SEs or CIs if available.

Heterogeneity was assessed using the I² statistic, which describes the true variation among studies as a percentage, and values were interpreted as low (25%), medium (50%), or high (75%) heterogeneity.²⁴  For those studies evaluating several muscles, only 1 muscle was included in the meta-analysis to avoid including the same study population multiple times (ie, double counting), which would have inherently increased the statistical weight.² 

Given high EMG data variability, the following data-extraction order was prioritized for the meta-analysis² : vastus lateralis > vastus medialis > rectus femoris > soleus > tibialis anterior > fibularis longus in the lower extremity and biceps brachii > triceps brachii > brachioradialis > forearm flexors in the upper extremity. Lower limb muscles were prioritized in studies involving both upper and lower limb exercises to minimize outcome variability.²  For studies performed with several muscle-action modes, the order was prioritized as isometric > isokinetic > concentric > eccentric, and priority was given to the open kinetic chain in studies of both open and closed kinetic chains. For those studies involving loads <20% of 1RM or MVC and several occlusion pressures, the highest BFR pressure (up to 80%) was considered. Moderate pressures (ie, 40% to 60% of total occlusion pressure) were prioritized in studies using several BFR levels and loads >20% of 1RM or MVC. In all analyses, multiple comparisons were included from several studies (eg, dynamic and isometric exercise, muscles of upper and lower limbs) to increase accuracy and thus the generalization of our results.²⁶ 

All statistical analyses were performed using the Cochrane Review Manager (RevMan) software. The α level was set a priori at P < .05.

A total of 1678 articles were retrieved from database searches. After removing duplicates, we submitted the remaining articles to title and abstract analyses, and 54 articles were considered eligible. Of these, 16 were excluded for the following reasons: no EMG assessment,^27–32  no free–blood-flow group,³³  no LL-BFR group,³⁴  lack of resistance training,^35,36  individual motor-unit assessment,³⁷  intermittent BFR,³⁸  collapsed data regarding LL-BFR and LL-RT,^39,40  lack of EMG assessment in the agonist muscle,⁴¹  and total BFR.⁴²  The remaining 39 studies,^{15,16,18,19,43–77}  published between 2006 and 2020, were included in the systematic review. Thirty-two studies^{15,16,18,19,43–70}  were included in meta-analyses (Figure 1). An overview of the studies is in Table 1.

Most of the included trials presented a high risk of performance bias and other bias. In addition, most of the studies presented an unclear risk of bias for random sequence generation, allocation concealment, blinding of outcome assessment, incomplete outcome data, selective reporting, and group similarity at baseline (Table 2).

A total of 548 participants were enrolled in the 39 included trials,^{15,16,18,19,43–77}  and data from 411 participants were included in the meta-analyses.

The short-term effects of BFR were verified in 33 studies,^* and the long-term effects of BFR on myoelectric activity were verified in 6 studies.^{68,69,72,73,75,77}  In the long-term studies, the training program duration ranged from 5 to 12 weeks, with a frequency of 2 or 3 times per week.

The training load ranged between 10% and 50% of 1RM or MVC for both LL-BFR and LL-RT and between 60% and 80% for HL-RT. The load was imposed using elastic resistance in 2 studies.^48,62  In 15 studies,^† repetitions were applied to failure, and in 24 studies,^‡ a predetermined number of repetitions was performed.

Myoelectric activity was assessed in the time domain in 31 studies,^§ in the frequency domain in 1 study,⁷⁶  and in both domains in 7 studies.^{19,44,53,69–71,74}  The surface EMG variables assessed in the time domain were the root mean square (RMS), integrated EMG (iEMG), peak of the EMG signal (EMGpeak), EMG amplitude, and average EMG, whereas mean power frequency (MPF), median frequency (Fmed), and central frequency (CF) were assessed in the frequency domain.

Based on 18 studies,^‖ LL-BFR presented a moderate effect in increasing muscle excitability when exercises were not performed to failure compared with LL-RT (SMD = 0.61; 95% CI = 0.34, 0.88; Z = 4.41; P < .001). Heterogeneity was low (I² = 39%, P = .04). When exercise was performed to failure (n = 10 studies^¶), no differences were observed in muscle excitation between LL-BFR and LL-RT (SMD = −0.14; 95% CI = −0.38, 0.10; Z = 1.17; P = .24). The I² statistic of 0% (P = .60) represented very low heterogeneity for this result (Figure 2). Regarding the frequency-domain values assessed during nonfailure protocols, in 2 studies^44,70  no differences in MPF were found, whereas in 2 studies, reduced CF⁷⁶  or MPF⁷⁴  was identified during LL-BFR compared with LL-RT. One study¹⁹  was conducted using a failure protocol, and no differences in Fmed were observed between LL-BFR and LL-RT (Table 1).

Changes in muscle excitation pre-exercise and postexercise were evaluated in 9 studies. Among the studies conducted with nonfailure protocols, in 6 studies,^{16,43,53,67,70,74}  differences in muscle excitability were not observed, whereas in 1 study,⁴⁴  a reduced RMS post–LL-BFR compared with LL-RT was identified. One study⁷¹  was conducted using a failure protocol, and a greater EMGpeak was observed post–LL-BFR compared with LL-RT. No postexercise differences were noted between LL-BFR and LL-RT regarding the frequency-domain measures, whether performed to failure⁷¹  or not^44,53,70,74  (Table 1).

Based on 6 studies,^{16,49,51,53,55,65}  HL-RT had a large effect in increasing muscle excitability compared with LL-BFR during nonfailure protocols (SMD = −1.13; 95% CI = −1.94, −0.33; Z = 2.76; P = .006). Heterogeneity was considerably high for this meta-analysis (I² = 76%, P < .001). When a failure protocol was used (n = 3 studies^46,56,57), HL-RT had a moderate effect on increasing muscle excitability compared with LL-BFR (SMD = −0.61; 95% CI = −1.01, −0.21; Z = 3.00; P = .003). The I² statistic of 0% (P = .98) represented very low heterogeneity (Figure 3). No authors evaluated frequency-domain values during LL-BFR compared with HL-RT.

In 2 studies^16,53  conducted using nonfailure protocols, no differences in muscle excitation post–LL-BFR compared with HL-RT were observed, and in 1 study,⁵³  no differences were seen in frequency-domain measures. We identified no studies that used a failure protocol and investigated either muscle excitability or the frequency-domain values post–LL-BFR compared with HL-RT (Table 1).

Based on 2 studies,^68,69  LL-BFR had a large effect in increasing muscle excitability compared with LL-RT during exercises (3–6 weeks) performed to failure (SMD = 1.09; 95% CI = 0.39, 1.79; Z = 3.07; P = .002). The I² statistic of 0% (P = .51) represented very low heterogeneity (Figure 4). Descriptively, in 1 study⁷⁵  using a failure protocol, high iEMG was demonstrated after 8 weeks of LL-BFR compared with LL-RT, and in 1 study⁷³  using a nonfailure protocol, similar RMS values occurred after 6 weeks of either LL-BFR or LL-RT training. In 1 study,⁶⁹  similar Fmed values were present after 6 weeks of LL-BFR or LL-RT performed to failure. No studies using nonfailure protocols and investigating long-term effects of LL-BFR compared with LL-RT on frequency-domain variables were identified (Table 1).

Given missing quantitative data, we compared LL-BFR and HL-RT via a descriptive and qualitative approach. Based on 3 studies, no differences in muscle excitability were observed after 6 to 12 weeks of LL-BFR or HL-RT performed to failure⁶⁹  or not.^72,77  In 1 study⁶⁹  conducted using a failure protocol, a higher Fmed in HL-RT than in LL-BFR was shown after 2 weeks of training. No studies conducted using a nonfailure protocol and investigating the long-term effects of LL-BFR compared with HL-RT on frequency domains were identified (Table 1).

In this systematic review, we aimed to examine the short- and long-term effects of LL-BFR on myoelectric activity and compare them with those of LL- and HL-RT. As previous researchers^18,19  found that exercising to failure might mediate the effects of LL-BFR and LL-RT on muscle excitability, we also intended to provide strong evidence regarding the true effects of fatiguing or nonfatiguing exercise protocols. Our findings indicated that LL-BFR acutely increased muscle excitability compared with LL-RT only during a nonfailure protocol, whereas muscle excitability was greater with HL-RT than LL-BFR during either a failure or nonfailure protocol. In addition, long-term LL-BFR may lead to higher muscle excitation compared with LL-RT. From a qualitative viewpoint, the studies were limited in number and had conflicting results; thus, the evidence is inconclusive about the short-term effects of LL-BFR versus LL- and HL-RT on muscle excitability pre-exercise and postexercise and the short- and long-term effects on EMG frequency-domain measures.

Our results demonstrated a higher level of muscle excitation favoring LL-BFR compared with LL-RT exercises using a matched number of repetitions but not when exercise was performed to failure. It has been hypothesized⁷  that the hypoxic muscular environment generated during LL-BFR can cause high metabolic stress levels and activate mechanisms for muscle growth induction. One of the hypothetical mechanisms is the increased fast-twitch fiber recruitment,^14,75  because these fibers may be more susceptible to exercise-induced cross-sectional area gains compared with slow-twitch fibers.⁷⁸  Although limitations exist regarding the inference of fiber-type recruitment using surface EMG, researchers³⁷  using high-density EMG and decomposition techniques have shown that LL-BFR facilitates the early recruitment of higher-threshold motor units compared with LL-RT.

Indeed, LL-BFR produces greater muscle strength and growth than does LL-RT if applied with repetition-matched protocols⁷⁹ ; however, this may be mitigated during exercises performed to volitional failure.^60,61  In this sense, and considering the reduced perceived exertion, discomfort, and soreness reported during a nonfailure protocol using LL-BFR,⁸⁰  our findings support prescribing a nonfailure protocol with LL-BFR as an effective approach to increase muscle excitability. The BFR can be useful in reducing the time to volitional failure during training performed to failure.⁸¹  This can save time and minimize overload because of the large number of repetitions performed during LL-RT.⁶¹ 

Frequency-domain values are generally used as indirect markers of muscle fatigue.¹⁹  Our qualitative analysis revealed that MPF and CF may or may not be reduced during nonfailure protocols using LL-BFR compared with LL-RT. This conflicting result may be due to methodologic differences, especially owing to the different muscles being studied. For example, frequency-domain measures may be reduced in the biceps brachii^74,76  but not in the vastus lateralis.^44,70  Differences in Fmed during LL-BFR compared with LL-RT were not observed in 1 study using a failure protocol,¹⁹  suggesting similar fatigue in both conditions during submaximal (45% MVC) efforts performed to task failure.

No differences in muscle excitability postexercise with matched repetitions were found in most included studies, whereas greater muscle excitability was noted immediately after LL-BFR in 1 study⁷¹  using a failure protocol. Furthermore, no differences were identified in frequency-domain values postexercise, regardless of muscle failure. To summarize, these results indicate similar muscle fatigue levels after LL-BFR and LL-RT.

Greater muscle excitability was observed during HL-RT compared with LL-BFR regardless of muscle failure, probably because higher loads demand greater motor-unit recruitment at exercise onset to produce higher force.¹⁰  It has been suggested that muscle excitability during LL-BFR with loads of 40% to 50% of 1RM can reach levels comparable with those demonstrated during HL-RT, whereas it can be lower during LL-BFR with reduced loads (approximately 20% of 1RM).^12,53,75  In fact, most included studies (8 studies^{16,46,49,51,53,55–57}  of 9 studies^{16,46,49,51,53,55–57,65}; Table 1) used loads between 15% and 30% of 1RM; thus, it remains unclear whether LL-BFR performed with loads of approximately 50% of MVC can result in muscle-excitation levels similar to those of HL-RT.

Given the previous knowledge regarding the indirect repercussions of high muscle-excitation levels on muscle-strength gains induced by resistance training,^12,13  our findings are in line with those of previous systematic reviews^1,7  in which researchers observed higher strength gains after HL-RT compared with LL-BFR. We did not locate any study comparing frequency-domain measures between LL-BFR and HL-RT. In this sense, it is unclear whether fatigue is different when these 2 conditions are performed or not performed to failure.

When the EMG signal was evaluated immediately postexercise, no differences were present in muscle excitation and frequency-domain values between LL-BFR and HL-RT performed with a matched number of repetitions, indicating similar fatigue levels. No studies evaluating muscle excitation and frequency-domain measures after LL-BFR and HL-RT performed to failure were identified.

In a recent meta-analysis, Centner and Lauber⁸²  determined that long-term LL-BFR training increased muscle excitability compared with LL-RT. We built on existing evidence¹⁹  that exercising to failure might be a potential influencing factor and examined subgroups of studies using failure and nonfailure protocols. Although studies using nonfailure protocols are scarce, our meta-analysis indicated that when LL-BFR was performed to failure, it increased muscle excitability compared with LL-RT. Long-term adaptations in muscle excitability also occurred after LL-BFR with a matched number of repetitions⁷⁵ ; however, evidence is still scarce and contradictory. In terms of underlying mechanisms, researchers^8,62  have suggested that increased muscle excitability after long-term LL-BFR training may be due to increased motor-unit recruitment or firing frequency.

Long-term adaptation of muscle excitability did not differ between LL-BFR and HL-RT performed or not performed to failure. This is intriguing because HL-RT immediately induces higher muscle excitability than does LL-BFR. In this sense, similar results would also be expected after a long training period, but this notion was not confirmed. Further investigation is necessary to explore differences between the immediate responses and long-term adaptations to resistance training using BFR.

Regarding frequency-domain values, limited data suggested a greater Fmed after 2 weeks of HL-RT compared with LL-BFR performed to failure. In addition, no studies were conducted using a matched number of repetitions.

Our study had several limitations. For example, proposing EMG variables as surrogate endpoints for both strength and hypertrophy may lead to misinterpretations because whether increased muscle excitability assessed using surface EMG predicts long-term adaptations is still unknown.^13,83  Also, we included studies that applied either individualized or arbitrary BFR pressures. This might have influenced the results because high occlusion pressures may be related to greater muscle activation.⁴⁹  Moreover, most of the included studies had a high or moderate risk of bias, and I² was very high in the comparison of LL-BFR versus HL-RT not performed to failure. This high risk of bias might partly have reflected the fact that participant blinding is often impossible in this research field. Given the high heterogeneity of studies and exercise protocols, further research is needed to allow better comparisons among studies and draw more solid conclusions. Finally, as several populations were mixed, future authors should address differences in muscle excitability between specific populations (eg, younger versus older populations, healthy versus patient populations). Therefore, our results must be interpreted while taking this context into account.

The topic of this present systematic review and meta-analysis is of high importance for sport practitioners and rehabilitation settings. Especially in orthopaedic rehabilitation, arthrogenic muscle inhibition and neural-drive alterations are frequent phenomena after anterior cruciate ligament reconstruction⁸⁴  and chronic ankle instability.⁸⁵  Hence, investigating potential rehabilitation and training strategies is vitally relevant. Although blood-flow restriction training has been shown to be effective in musculoskeletal rehabilitation,^5,86  data regarding underlying mechanisms are scarce. Here, we demonstrated that adding BFR to low-load exercise regimens increased EMG amplitudes during nonfailure protocols. However, HL-RT regimens seemed to be superior to LL-BFR. Enhancements in neural drive and muscle excitability might be beneficial for sports in which ballistic and reactive contractions are necessary and for populations with an increased risk of falling to improve postural control.⁸⁷ 

Our results provide evidence of greater muscle excitability during acute LL-BFR compared with LL-RT using nonfailure protocols. Muscle excitability appeared to be greater during HL-RT than during LL-BFR, regardless of muscle failure. Given the scarce and conflicting evidence regarding long-term adaptations, future longitudinal studies should be done to consider volitional failure as a prescription variable that interferes directly with myoelectric activation and indirectly with strength and muscle-mass gains after LL-BFR training.

Mikhail Santos Cerqueira, MSc, PT, and Daniel Germano Maciel, MSc, PT, thank Coordination for the Improvement of High Level (CAPES, Brazil), finance code 001, for the scholarship concession.