Effects of exercise training with blood flow restriction on vascular function in adults: a systematic review and meta-analysis

Search strategy

The review protocol was registered with the International Prospective Register of Systematic Reviews (PROSPERO, CRD42019147408) and reported according to the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) (Moher, 2015). Five electronic databases (PubMed, Web of Science, Scopus, Medline and EmBase) were searched for studies published from inception to 31^st August 2019 and updated on 19^th August 2020, with no language restriction imposed. The following combination of terms were used within the search strategy: (“blood flow restriction” OR “blood flow occlusion” OR “vascular occlusion” OR “kaatsu training”) AND (“vascular function” OR “endothelial function” OR “arterial stiffness” OR “pulse wave velocity” OR “flow-mediated dilatation” OR “VEGF1” OR “nitric oxide”) (See Table S1 for example search strategy). The reference lists of relevant studies identified in the screening process were screened for any additional relevant studies.

Eligibility criteria

Studies were eligible for inclusion in this review if they: (a) were a primary study published as full text in a peer-reviewed journal; (b) were a controlled trial with or without randomization (parallel groups, cross-over included), with at least one group including BFRE and, at least one group including non-BFRE (exercise of any form); (c) were conducted in adults (18 years and over with or without chronic health conditions); (d) applied blood flow restriction (BFR) using an external device to promote external pressure in the proximal part of a limb (e.g., cuff, elastic band); and (e) reported at least one outcome related to vascular function (e.g., flow-mediated dilatation (FMD), pulse-wave velocity (PWV), nitric oxide). Studies were excluded if they: (a) were single group pre-post studies comparing baseline outcomes with outcomes measured during or post-BFRE without a non-BFRE group comparator; or (b) used other forms of hypoxic training such as hyperbaric chamber or high-altitude training as the intervention group.

Selection process

References identified from database searches were imported to systematic review software (Covidence systematic review software, Veritas Health Innovation, Melbourne, Australia) and exact duplicates removed. Title and abstract of references were initially screened for eligibility by two reviewers independently (EPN and MW). Where studies met eligibility or could not confidently be excluded, full text versions were accessed and screened by two reviewers independently (EPN and HB). Differences between reviewers were resolved by discussion and when consensus could not be reached, a third reviewer (MTW) was consulted. Hand searching of reference lists of included papers was undertaken to identify additional eligible studies.

Data extraction

A data extraction template was developed and piloted a priori. Data extraction was performed by two reviewers independently (EPN and HB). Data extracted from studies included the following domains:

(a) Publication demographics (publication year, authors, country of data collection);

(b) Study details (design, sample size, participant age (mean, SD) and gender, health status (healthy or clinical condition));

(c) Exercise training regimen for intervention and control groups (protocol duration, exercise intensity, frequency and type, blood flow restriction protocol (intervention group), blood flow restriction pressure (intervention group));

(d) Primary outcomes: mean and standard deviation (SD) for vascular function outcomes pre- and post-intervention for the BFRE and non-BFRE groups, which could include but were not limited to: vascular structure (e.g., ankle-brachial index, arterial diameter, pulse-wave velocity), endothelial function (e.g., flow-mediated dilatation, resting blood flow, reactive hyperemia blood flow) and biomarkers (e.g., catalase, vascular endothelial growth factor, nitric oxide). For studies where the results for the primary outcomes were presented in figures, WebPlotDigitizer was used to extract data (Rohatgi, 2019);

(e) Secondary outcomes: adverse events (any unpleasant or unintended event, sign, symptom reported by authors as related to the intervention (Australian National Health and Medical Research C, 2018)), adherence (percentage of planned sessions attended (Sohanpal et al., 2012)) and attrition (proportion of randomized participants analyzed (Sohanpal et al., 2012)). Corresponding authors were contacted to seek for any missing primary or secondary outcomes.

Study quality—appraisal of methodological bias

The methodological quality of the studies was assessed using the Tool for the Assessment of Study Quality and Reporting in Exercise (TESTEX) scale (validity and reliability intra-class correlation coefficient >0.91 (Smart et al., 2015)). This instrument consists of two sections (study quality and reporting). Standardized instructions guided scoring (minimum of 0 and maximum of 15 points), where higher scores indicated better study quality (0–5 points) and reporting (0–10 points) (Sohanpal et al., 2012). Assessment was performed by two reviewers independently (EPN and JP) for all eligible studies. Studies were not excluded based on score for methodological quality as this has been shown to reduce precision and increase bias on meta-analysis results (Stone et al., 2019; Higgins et al., 2011).

Data management and analysis

Study design and population

The majority of included studies were controlled trials (13 randomized controlled trials (RCT), 10 randomized cross-over trials, two non-randomized controlled trials and a case report; see Table S2). There were 15 (58%) studies that evaluated BFRE versus one other group, six studies had two comparator groups and five studies had three comparator groups (See Table S2). Across the 26 studies, 472 individuals (303 males, 169 females, aged 18 to 91 years) were included. The majority of studies (n = 23, 88%) were in non-clinical populations (total n = 431, 284 males, 147 females, aged 18 to 81 years). Three studies recruited participants with chronic conditions (total n = 41; 19 males, 22 females; aged 58 to 91 years), with each study representing a different clinical condition (coronary artery disease, hypertension, sarcopenia). Overall, methodological quality was moderate (median 9.5) with scores ranging between 6–13 (Details presented in Table S3).

Comparators

The most frequent comparators to the active intervention (low-intensity exercise with BFR) were exercise training without BFR at low (n = 17, 65%) or higher (n = 8, 31%) exercise intensities. Two studies included four groups (with and without BFR at low- and moderate-intensity exercise). Single studies compared different (i) restriction pressures (40% and 80% of maximum occlusion pressure) or (ii) BFR position (upper and lower limbs) with low and high intensity exercise to low and high intensity exercise without BFR. The study that used neuromuscular electrical stimulation (NMES) compared the same intensity of maximum voluntary contraction (MVC) with and without BFR.

Blood flow restricted exercise protocols

Primary outcomes

Table 1 presents a summary of outcomes for Vascular structure, Endothelial function and Biomarkers.

Meta-analysis

Nine studies were sufficiently homogenous (design, training protocol and population) for inclusion in the meta-analysis (Ozaki et al., 2013; Patterson & Ferguson, 2010, 2011; Shimizu et al., 2016; Clark et al., 2011; Fahs et al., 2012; Mouser et al., 2019; Yasuda et al., 2015a; Yasuda et al., 2015b). All studies included in the meta-analysis were in non-clinical populations and explored the effects of BFR with isotonic dynamic resistance training for four weeks or more. Separate models were run comparing the effects of low intensity BFRE to high intensity non-BFRE on vascular structure, and low intensity BFRE to low intensity non-BFRE on endothelial function.

Low intensity BFRE versus high intensity non-BFRE—Vascular structure

Four studies reported outcomes for vascular structure (vascular compliance (n = 3), (Ozaki et al., 2013; Fahs et al., 2012; Mouser et al., 2019); and PWV (n = 1) (Clark et al., 2011)) after four or more weeks of dynamic resistance training (low-intensity BFRE versus high-intensity non-BFRE) (Fig. 1). No significant difference was observed between low intensity BFRE and high intensity non-BFRE (SMD -0.24, 95% CI [−1.08 to 0.59], p = 0.57) (Fig. 1). Moderate and significant heterogeneity was observed (I² = 73%, p = 0.01).

Low intensity BFRE versus low intensity non-BFRE—Endothelial function

Five studies provided sufficient data for meta-analysis of the effect of four or more weeks of low intensity dynamic resistance exercise training with and without BFR on endothelial function (Fig. 2). Specific endothelial outcomes were FMD (n = 2) (Yasuda et al., 2015a, 2015b), reactive hyperemia blood flow (n = 2) (Patterson & Ferguson, 2010, 2011) and reactive hyperemia index (n = 1) (Shimizu et al., 2016). Meta-analysis estimated low-intensity dynamic resistance training with BFR had a significant positive difference compared to low intensity exercise without BFR (SMD 0.76, 95% CI [0.36 to 1.14], p < 0.001) for endothelial function (Fig. 2). Heterogeneity was considered low (I² = 7%) and not significant (p = 0.36).

Meta-regression

Findings from meta-regression analyses are summarized in Table S5 (Lopes et al. (2019) excluded as case study). For vascular structure, 15 studies were included within meta-regression models reflecting four vascular structure outcomes: pulse-wave velocity (PWV) (n = 5) (Credeur, Hollis & Welsch, 2010; Clark et al., 2011; Credeur et al., 2019; Fahs et al., 2014; Figueroa & Vicil, 2011), vascular compliance (n = 5) (Karabulut, Karabulut & James, 2018; Ozaki et al., 2013; Fahs et al., 2012; Mouser et al., 2019; Sardeli et al., 2017), cardio-ankle vascular index (n = 3) (Yasuda et al., 2015a; Yasuda et al., 2015b; Yasuda et al., 2016) and arterial diameter (n = 2) (Paiva et al., 2016; Renzi, Tanaka & Sugawara, 2010). No significant associations were identified between BFRE protocol factors and effect sizes (all p-values >0.05).

For endothelial function, 13 studies were included in the meta-regression reflecting three outcomes: FMD (n = 8) (Paiva et al., 2016; Renzi, Tanaka & Sugawara, 2010; Hunt, Walton & Ferguson, 2012; Yasuda et al., 2015a; Yasuda et al., 2015b; Credeur et al., 2019; Ramis et al., 2020; Kambic et al., 2019), reactive hyperemia blood flow (n = 4) (Patterson & Ferguson, 2010, 2011; Credeur et al., 2019; Fahs et al., 2014) and reactive hyperemia index (n = 1) (Shimizu et al., 2016). Study design was significantly associated with effect size, with randomized cross-over trials associated with smaller effect sizes compared to control trials (regression estimate (ß) of −2.91; 95% CI [−5.40 to −0.42]; p = 0.02). For exercise modality, compared to handgrip training, treadmill walking (ß = −2.62; 95% CI [−4.18 to −1.06]; p = 0.001) and dynamic resistance training (ß = 1.64; 95% CI [0.83–2.45]; p < 0.001) were associated with significantly smaller and larger effects on endothelial function, respectively. Upper limb only exercise was associated with smaller effect sizes (ß = −2.22; 95% CI [−4.23 to −0.19]; p = 0.03) compared to exercise of the lower and upper limbs combined. For the number of exercise sets, compared to studies of ≤2 sets, exercise protocols with three (ß = 1.66; 95% CI [0.21–3.11]; p = 0.02) and four (ß = 1.54; 95% CI [0.03–3.05]; p = 0.05) sets were associated with larger effect sizes, while protocols with five or more sets (ß = 2.5; 95% CI [−4.69 to −0.30]; p = 0.03) were associated with smaller effect sizes. Meta-regression analysis for length of intervention was performed using ‘single session’ studies as the reference for duration, with durations explored as a continuous variable (number of weeks, p > 0.05) or categorical variable of ≥4 weeks (e.g., all studies with 4 weeks or more grouped in one category). Compared to a single session, interventions of ≥4 weeks were associated with larger effect sizes (ß = 1.92; 95% CI [0.39–3.46]; p = 0.01).

Secondary outcomes

Table S6 presents a summary of adverse events, adherence, and attrition rates. The majority of studies (n = 18, 69 %) did not report the presence or absence of adverse events. Seven studies explicitly reported an absence of adverse events (n = 7, 27%). A single study reported an initial sensation of leg numbness in two participants once they commenced BFRE, which did not persist beyond the first minute of exercise (Barili et al., 2018) (2 out of 16 hypertensive older women). Adherence and attrition were reported in 11 (42%) and six (23%) studies, respectively. Adherence was reported or calculated as 100% for 11 studies and dropout rates ranged from 5% to 40%.

BFRE and vascular structure

Our meta-analysis did not identify significant differences between BFRE and non-BFRE for arterial stiffness outcomes. Increased sympathetic nervous system activity is a common concern with BFRE (Da Cunha Nascimento, Schoenfeld & Prestes, 2020). The reduction of blood flow to the muscle increases vascular peripheral resistance and metabolite concentration, which may trigger baro- and chemoreceptors that stimulate sympathetic outflow (Renzi, Tanaka & Sugawara, 2010). Domingos and Polito (2018)’s meta-analysis reported BFRE to cause greater increases in blood pressure than low intensity traditional resistance training (Domingos & Polito, 2018), indicating exercise-induced increases in sympathetic activity, which might be expected to increase arterial stiffness and negatively affect vascular structure. The same meta-analysis (Domingos & Polito, 2018) and previous randomized trials (Cezar et al., 2016; Araújo et al., 2014) have reported a more pronounced post-exercise hypotensive effect for BFRE compared to high- and low intensity traditional resistance training, which is indicative of reduced sympathetic activity (Cristina-Oliveira et al., 2020). Despite greater sympathetic activity with BFRE compared to low intensity exercise without BFR, the post-exercise reduction of sympathetic outflow associated with BFRE could potentially act as a protector of the vascular system.

BFRE and endothelial function

In healthy individuals, our meta-analysis indicated that four or more weeks of low intensity dynamic resistance exercise with BFR was associated with greater improvements in endothelial function compared to low intensity non-BFRE. Increased shear stress is hypothesized to be a modulator of chronic endothelial function changes following exercise (Pedralli et al., 2018; Paiva et al., 2016). Improvements in FMD following traditional moderate-to-high dynamic resistance exercise have been proposed to be associated with the process of intermittent blood flow in the small muscular arteries resulting from muscle contraction, increasing shear stress in the artery walls and promoting release of vasodilator factors (i.e., NO, VEGF, GH) (Paiva et al., 2016). It is possible that hypoxic conditions created during prolonged exercise with blood flow occlusion and/or the increased shear stress following release of the cuff, increase the availability of vasodilator factors consequently improving endothelial function.

BFRE and vascular function biomarkers

Few studies included in this review (n = 10; 35%) reported VF biomarker outcomes (Ozaki et al., 2013; Shimizu et al., 2016; Larkin et al., 2012; Natsume et al., 2018; Barili et al., 2018; Boeno et al., 2018; Yasuda et al., 2015a, 2015b; Yasuda et al., 2016; Ramis et al., 2020). Within these studies, seven different outcomes were reported, which in included outcomes not unique to vascular function (i.e., growth hormone (Natsume et al., 2018)). Only four of these studies reported between group differences for VF biomarkers (Shimizu et al., 2016; Natsume et al., 2018; Barili et al., 2018; Boeno et al., 2018) and, while all four favored BFRE, these studies differed in design, restriction pressure, exercise regimen, exercise intensity and assessed different outcomes (growth hormone, von Willebrand factor, catalase and superoxide dismutase) making it difficult to draw conclusions on the overall effects of BFRE on vascular function biomarkers. Given the variety of outcomes observed in this review, and the heterogeneity of study designs/protocols, it is not clear which biomarkers should be considered as priorities for studies investigating BFRE and vascular function.

Which characteristics of BFRE protocols were associated with vascular function or structure effects sizes? Findings of the meta-regression

Our meta-regression results should be interpreted with caution due to the small number of studies (total k = 15 for vascular structure and k = 13 for endothelial function, with fewer common studies for each specific outcome within the broader domains), high heterogeneity between study protocols, and recognition that within each bivariate model, effect sizes for each variable and differences between strength of association are compared with a reference variable, negating identification of the causal influence of different predictors (Higgins et al., 2019). In addition, only four (16%) studies reported a sample size estimate, making it difficult to determine whether the remaining studies (n = 22; 84%) were adequately powered. Underpowered studies may provide inflated effect sizes and reduce the chance of detecting a true difference (Button et al., 2013; Turner, Bird & Higgins, 2013). With these forewarnings in mind, a number of protocol ‘signals’ were evident for endothelial function effects (but not vascular structure).

Intervention duration as a categorical, but not a continuous variable, was a significant predictor of effect size. Endothelial function measures are prone to change after a single bout of exercise (Ashor et al., 2015), the degree of which appears to differ between exercise type (Ashor et al., 2015). As such, the time frame required for physiological changes to endothelial function in response to chronic exercise varies. It is likely that other BFRE protocol factors (e.g., approach to determine pressure, cuff width) influence the relationship between intervention duration and endothelial function, which could only be detected in a multivariate analysis.

Consistent with known exercise training principles (Schoenfeld, 2013), we found exercise volume and mode (engaging larger muscle groups) to be important for endothelial function outcomes. Previous meta-analysis on non-BFRE (i.e., traditional exercise training) identified higher exercise volumes, compared to lower exercise volumes, resulted in greater improvements in skeletal muscle hypertrophy (Schoenfeld, Ogborn & Krieger, 2017), potentially caused by an increased release of neurohumoral factors (i.e., GH, nitric oxide, vascular endothelial growth factor). These same neurohumoral factors are known to modulate endothelial function (Schoenfeld, 2013). It is, therefore, possible that higher training volumes will also impact vascular outcomes. Within our meta-regression, there were also variables signaling potential harm – albeit from a single study. The association between training at five or greater sets per exercise and smaller effect sizes for endothelial function may reflect the concept that elevated training volumes can produce higher inflammatory responses (Calle & Fernandez, 2010) and, consequently, decrease endothelial function. This is in agreement with a recent review of BFRE methodology, application and safety which recommended BFR training use dynamic resistance exercise of four sets per exercise (Patterson et al., 2019).

Within our review, reported cuff widths ranged between 3 to 23 cm (n = 18) and the most common approach reported for determining restriction pressure was use of a predetermined, non-individualized pressure (n = 13) with systolic blood pressure (n = 10) and arterial occlusion pressure (n = 3) accounting for the remaining approaches. The approach used to determine restriction pressure (e.g., predetermined non-individualized, systolic blood pressure or LOP) did not significantly predict of effect size for arterial structure or endothelial function. The review published by Patterson et al. (2019) recommended using a percentage of an individual’s LOP. The recency of this recommendation (published May 2019) is likely to explain why the majority (92%) of studies did not use LOP to prescribe restriction pressure. Limitations of alternative approaches to determining BFRE restriction training pressures include inter limb/individual variation in limb circumferences necessitating different cuff pressures and inconsistent associations between blood pressure and the pressure required to achieve occlusion (Loenneke et al., 2012). Higher restriction pressures (i.e., non-individualized 220 mmHg, which is 100 mmHg higher than normal SBP) also appear to increase discomfort and cardiovascular responses (Jessee et al., 2017), which may increase the risk of adverse vascular events (Loenneke et al., 2012). Future studies in BFRE and vascular function are more likely to individually prescribe training pressure using a percentage of LOP.

Blood pressure responses following BFRE have been explored in prior meta-analysis (23 studies included in the analysis of normotensive and hypertensive individuals (Domingos & Polito, 2018)), with hypertensive participants showing higher blood pressure changes. There is very little data available on the effects of BFRE in the vascular function of older individuals with chronic conditions. Our review included 41 individuals (of the total n = 472) with chronic conditions, and the review by Domingos & Polito (2018) included 95 (of the total n = 325). As populations living with chronic conditions often present with increased arterial stiffness and endothelial dysfunction (Chen et al., 2015), it is likely that they will have different vascular function responses to BFRE than healthy adults. Notably, the majority of studies included within this review excluded participants with comorbidities such as peripheral vascular disease (assessed by the ankle-brachial index) (Loenneke et al., 2012). The findings of this review concerning the effects of BFRE on vascular function should therefore not be generalized to older populations living with chronic cardiovascular comorbidities.

Secondary outcomes

In a survey (Patterson & Brandner, 2017) of 115 practitioners who used BFRE, 39% (n = 60) reported at least one side effect observed in their practice. Whereas in Hughes et al. (2017)’s review of BFRE and our review, 50% and 69% of studies (respectively) made no mention of adverse events. In these studies, it is not clear if adverse event did not occur or if they were not assessed or reported. The potential underreporting of either presence or absence of adverse events makes it difficult to determine the safety of BFRE and transparent reporting should be considered a priority in future BFRE studies.

Strengths of this review include a comprehensive search of major databases related to the review topic, with no language or time restrictions, and a rigorous approach to conducting the review and reporting of results. Due to heterogeneity of exercise regimens and outcomes in the studies included, it was not possible to perform adjustment for potential confounding factors in the meta-regression. Publication bias was not assessed as the funnel plot tests to detect asymmetry are compromised in meta-analysis with small numbers (<10) of studies (Higgins et al., 2019). Translation of our results to clinical populations is limited as the majority of included studies were in healthy individuals.

Key findings:

• The evidence regarding the effects of blood flow restricted exercise (BFRE) training in vascular function is small and heterogeneous.

• Our results suggest that BFRE training may have a positive effect on endothelial function and that BFRE protocol factors (e.g., mode and duration of exercise) are associated with larger effect sizes for vascular function outcomes.

• It is necessary to report adverse events and include vascular function outcomes in future BFRE training studies to clarity its safety.