Comparative effects of aerobic training modalities with and without blood flow restriction on maximal and peak oxygen uptake in adults: A systematic review and network meta-analysis Authors: Hangxin Chen, Feng Xu, Ziqi Fan, Yangyang Liu, Dongxu Gao Published: 2026 ABSTRACT BackgroundMaximal oxygen uptake (VO2max) and peak oxygen uptake (VO2peak) are key indicators of cardiorespiratory fitness. Blood flow restriction (BFR) training may augment the metabolic stimulus elicited by aerobic exercise (AE) across a range of exercise intensities in adults. However, the comparative effects of BFR-based AE strategies and conventional AE on VO2max/VO2peak remain uncertain. This network meta-analysis (NMA) compared the relative effects, at the group level, of BFR-based AE strategies and conventional exercise interventions on VO2max/VO2peak in adults across different exercise modalities, intensity levels, and comparator conditions, thereby providing a context-specific reference for understanding the potential effects of these exercise strategies on aerobic capacity.MethodsSix databases (PubMed, Web of Science, Embase, Cochrane Library, Scopus, and SPORTDiscus) were searched through January 21, 2026 for randomized controlled trials involving adults and reporting VO2max or VO2peak. A Bayesian NMA was performed using the gemtc package in R v4.4.1. Relative effects were summarized with league tables, and intervention rankings were estimated using the surface under the cumulative ranking curve (SUCRA).ResultsThirty-four studies including 1042 participants were included. Overall, high-intensity interval training combined with BFR (HIIT + BFR), low-intensity continuous training combined with BFR (LICT + BFR), and moderate-intensity continuous training (MICT) ranked highest. Subgroup analyses showed that HIIT + BFR ranked highest in 4–6-week interventions, trained participants, and running-based protocols. LICT + BFR ranked highest in untrained participants, walking-based protocols, and low-to-moderate-intensity exercise. MICT ranked highest in 8–12-week interventions, male participants, cycling-based protocols, and trained participants. Low-intensity interval training alone showed a consistently low ranking.ConclusionAt the group level, BFR combined with AE may improve VO2max/VO2peak in some intervention settings, while MICT remains an effective conventional training strategy. However, given the limited comparative evidence, these findings should be interpreted cautiously, and further high-quality studies are needed to inform protocol-specific recommendations. Background Maximal oxygen uptake (VO2max) is the maximum amount of oxygen that the body can take up, transport, and utilize during exercise with increasing load. It is regarded as a comprehensive phenotype of the oxygen utilization capacity of the cardiorespiratory, circulatory, and musculoskeletal systems, and is one of the most classic and important indicators for evaluating aerobic capacity.1 In the context of endurance exercise, VO2max constitutes the upper limit of sustained aerobic energy production and, together with the lactate threshold and exercise economy, reflects an individual's potential for endurance performance.2 More recently, fatigue resistance, or durability, has been proposed as a potential “fourth dimension” of endurance performance in addition to these classical determinants.3 Studies have revealed a close link between better cardiorespiratory fitness (CRF) and lower risks of all-cause mortality, cardiovascular events, and various chronic diseases; in certain contexts, its predictive value even surpasses that of some traditional risk factors 4, 5, 6. VO2max is typically improved through sustained endurance training at moderate to high intensities, threshold training, and high-intensity interval training (HIIT) 7, 8, 9. Studies have consistently demonstrated that both sustained endurance training and HIIT can significantly improve VO2max, and HIIT may yield greater gains in certain situations; however, this superiority is moderated by various factors (such as training volume, duration, and the subject's baseline fitness level) 7, 8, 9. However, the relative effects of HIIT and other aerobic training approaches may depend on intervention duration. Early physiological adaptations observed during short-term HIIT do not necessarily imply a sustained advantage after longer-term training, because long-term training adaptations involve complex central and peripheral mechanisms. Therefore, intervention duration should be considered an important contextual factor when interpreting the relative effects of different aerobic training strategies on VO2max/VO2peak.10,11 These traditional protocols are typically associated with high mechanical load, metabolic stress, and time costs. For in-season athletes, additional high-intensity aerobic exercise (AE) sessions often compete with sport-specific training, recovery requirements, and injury management. For individuals in rehabilitation or those with low baseline fitness, the feasibility of such interventions may be compromised by high impact, intense subjective fatigue, and low tolerance.9,12 In recent years, blood flow restriction (BFR) training has been applied to increase muscle strength and induce hypertrophy while using relatively low external loads. By restricting arterial inflow and venous return, BFR can augment local metabolic and systemic physiological responses during AE.13,14 These responses may elicit a substantial internal training stimulus at a relatively lower external load than may otherwise be required during conventional training to produce a comparable physiological demand. External load is typically reflected by exercise speed, power output, or accumulated mechanical work. The potential advantage of BFR therefore lies in its capacity to promote aerobic adaptations while reducing the external load required to generate a substantial physiological stimulus. These adaptations may be mediated by reduced local oxygen availability, promoting metabolite accumulation, increasing high-threshold motor unit recruitment, amplifying molecular signaling pathways related to mitochondrial biogenesis (such as the adenosine monophosphate-activated kinase [AMPK]/peroxisome proliferator-activated receptor gamma coactivator-1α [PGC-1α] pathway), and inducing vascular remodeling and improvement in oxygen delivery capacity 15, 16, 17, 18. In some exercise contexts, a lower external load may also reduce musculoskeletal load, suggesting potential value of BFR in rehabilitation and in-season training load management. For elite athletes, BFR may also serve as an adjunctive training stimulus when further increases in conventional training volume or intensity are difficult to tolerate. Nevertheless, a lower external load should not be equated with lower perceptual discomfort or physiological strain. Particularly in individuals with low fitness levels and those undergoing rehabilitation, the feasibility and tolerability of BFR may be influenced by cuff-related factors, individual health status, pressure prescription, familiarization, and the availability of appropriate professional supervision 12, 13, 14, 15, 16, 17, 18. Evidence from controlled studies suggests that low-intensity AE combined with BFR may improve VO2max, although the interpretation of this finding depends on the comparator condition and study design. In a 2-week randomized controlled trial (RCT) involving collegiate basketball players, Park et al. reported a significant group × time interaction, with BFR walking training yielding greater improvements in VO2max than matched walking training performed without BFR.19 Similarly, Held et al. found that, in elite rowers, 5 weeks of low-intensity rowing combined with BFR produced greater increases in VO2max than the identical rowing program performed without BFR.20 In contrast, de Oliveira et al. compared four interval-training strategies—low-intensity training with BFR, low-intensity training without BFR, HIIT, and HIIT combined with BFR—and found that VO2max improved from baseline in the low-intensity BFR, HIIT, and HIIT + BFR groups, but no statistically significant between-group differences were observed.21 Recent studies have attempted to combine BFR training with low-intensity continuous training (LICT) or high-intensity interval training (HIIT) to explore improvements in cardiopulmonary function indicators, including VO2max, through these combined intervention protocols. However, existing studies exhibit significant variations in intervention parameters (e.g., pressure, duration, and frequency), and most studies have small sample sizes, resulting in inconsistent conclusions. Given these limitations, this review employed a network meta-analysis (NMA) to compare the effects of different AE protocols on VO2max/VO2peak. To improve comparability across studies, exercise intensity was standardized a priori into three categories adapted from the American College of Sports Medicine (ACSM) relative-intensity criteria: low intensity (<40% of heart-rate reserve [HRR] or oxygen-uptake reserve [VO2R], or < 64% of maximal heart rate [HRmax]), moderate intensity (40%–59% of HRR or VO2R, or 64%–76% of HRmax), and high intensity (≥60% of HRR or VO2R, or ≥ 77% of HRmax).22 Multiple intervention protocols, including LICT + BFR, HIIT + BFR, HIIT, and LICT, were incorporated into a single analytical framework to synthesize direct and indirect evidence, estimate their relative effects using mean differences (MDs), and rank their comparative efficacy. This approach extends beyond traditional pairwise meta-analysis by enabling the simultaneous comparison of multiple interventions and provides a more comprehensive evaluation of the relative effects of BFR-based AE strategies across different exercise modalities and comparator conditions. The findings may therefore offer group-level comparative evidence to inform the design of exercise training and rehabilitation programs, while acknowledging that the relative effectiveness of different strategies may vary according to the specific exercise and application context. Methods This study strictly adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Network Meta-Analyses (PRISMA-NMA) guidelines.23 The complete PRISMA-NMA checklist is provided in Supplementary Table S1. Additionally, this study was registered in the International Prospective Register of Systematic Reviews (PROSPERO) under registration number CRD420261278478. Inclusion and exclusion criteria Based on the Population, Intervention, Comparison, Outcomes, and Study design (PICOS) principles, comprehensive inclusion criteria were established at the initial stage of the study: (i) Population: Adults with a clearly defined training background or physical activity level who were capable of completing the training intervention and outcome assessments; (ii) Interventions: Any exercise regimen combined with BFR, including LICT + BFR, low-intensity interval training (LIIT) + BFR, moderate-intensity interval training (MIIT) + BFR, and HIIT + BFR; (iii) Comparators: Eligible studies were required to include at least one BFR-based AE arm and at least one of the following comparator conditions: (1) the same AE regimen performed without BFR; (2) another non-BFR AE strategy differing in exercise modality or intensity; (3) another BFR-based AE strategy differing in exercise modality or intensity; or (4) CON, defined as a no-exercise control group; (iv) Outcome measures: Eligible studies were required to report pre- and post-intervention VO2max and/or VO2peak values and provide sufficient data to calculate changes from baseline. Because VO2max and VO2peak are closely related indicators of maximal aerobic capacity and have broadly similar physiological interpretations, they were treated as measures of the same outcome construct and combined in the analysis; (v) Study design: RCTs. Exclusion criteria included: (i) Non-original studies, including reviews and meta-analyses; (ii) Case reports, study protocols, etc.; (iii) Animal studies, in vivo/in vitro experiments; (iv) Conference abstracts, letters, guidelines, commentaries, editorials, and expert opinions; (v) Studies with no available access to the full text, duplicate publications, or studies with no extractable valid data. Literature search To comprehensively search for evidence on the efficacy of BFR-related exercise interventions for the outcomes of interest, we systematically searched PubMed, Web of Science Core Collection, Cochrane Library, Embase, EBSCO (SPORTDiscus), and Scopus. The search employed a combination of subject headings and free-text terms. The primary search terms comprised “Blood Flow Restriction Therapy” OR “Blood Flow Restriction Training” OR “Blood Flow Restriction Exercise” OR “BFR Therapy” OR “BFR Therapies” OR “blood flow restriction training” OR “BFR exercise” OR “BFR training” OR “blood flow restriction” OR “blood flow restriction exercise.” The search strategy was adjusted according to the search rules of each database. The literature search was completed on January 21, 2026. To minimize search bias, we applied no additional restrictions. The search strategy is detailed in Supplementary Table S2. Literature screening The retrieved records were imported into EndNote X9 (Clarivate Analytics, Inc., Philadelphia, PA). Two independent reviewers (Hangxin Chen and Feng Xu) conducted literature screening. First, automatic duplicate checks were performed via EndNote to ensure that the same publication did not appear multiple times in the software. Subsequently, the articles were screened through title and abstract assessment based on the inclusion and exclusion criteria. Full texts of articles that passed the initial screening were retrieved and further screened. Detailed reasons for exclusion were recorded. For any discrepancies between the two reviewers, a third independent reviewer (Dongxu Gao) made the final determination. Data extraction Two reviewers (Ziqi Fan and Yangyang Liu) extracted data from the included studies into a customized data extraction form created in Excel (version 16.93, Microsoft Corporation, Redmond, WA) and finalized before the full-text screening. The two reviewers independently extracted data regarding author, publication year, region or country, study characteristics, sample size, gender distribution, age, participant training status, training protocols, and outcome measures. Participant training status was retrospectively determined at the study level using an adapted version of the participant classification framework proposed by McKay et al. Participants were categorized as trained when the original study described them as meeting the criteria for Tier 2 (trained/developmental) or above, as indicated by regular, structured sport-specific training and/or a defined competitive or performance level. General physical activity alone was not considered sufficient to meet the criteria for trained status.24 For any discrepancies arising from data extraction, a third researcher (Dongxu Gao) made the final decision. If data were missing or presented solely in graphical form, the researchers contacted the authors to request the necessary data. If the contact was unsuccessful and the data remained available only in graphical form, WebPlotDigitizer 4.1 was utilized for data extraction. Any studies for which missing data could not be obtained were excluded from the final analysis. Mean, standard deviation, and sample size data before and after intervention were extracted for each group. Exercise intensity classification Exercise intensity was classified a priori using a three-level framework adapted from the ACSM relative-intensity criteria. Low intensity was defined as < 40% of HRR or VO2R, or < 64% of HRmax; moderate intensity was defined as 40%–59% of HRR or VO2R, or 64%–76% of HRmax; and high intensity was defined as ≥ 60% of HRR or VO2R, or ≥ 77% of HRmax.22 The ACSM very-light and light categories were merged into the low-intensity category, whereas the vigorous and near-maximal to maximal categories were merged into the high-intensity category. When multiple intensity indicators were reported, reserve-based measures, including HRR and VO2R, were preferred because they account for individual resting values. When reserve-based measures were unavailable, HRmax or the exercise-intensity indicator reported in the original studies was used. For studies reporting intensity using VO2max, VO2peak, PPO, Pmax, or other load-based indicators, classification was based on prespecified indicator-specific thresholds. Two reviewers independently verified the classification of each intervention, and disagreements were resolved through discussion or consultation with a third reviewer. Quality assessment Two researchers (Hangxin Chen, Feng Xu) assessed the risk of bias in the included RCTs utilizing the Cochrane risk-of-bias tool version 2 (RoB 2). Each study was assessed for 5 domains, namely selective reporting of results, outcome measurement, missing outcome data, deviations from intended interventions, and randomization process. Each domain was classified as low risk, some concerns, or high risk. If all domains were classified as low risk, the overall risk was rated as low; if at least one domain was classified as some concerns and no domains were classified as high risk, the overall risk was rated as some concerns; if any domain was classified as high risk, the overall risk was rated as high. Any discrepancies were addressed primarily through discussion; if consensus could not be reached, a third reviewer (Dongxu Gao) made the final decision. Data analysis A Bayesian NMA was conducted using the gemtc package in R v4.4.1 and the JAGS program. The model was constructed using four Markov chains with a thinning interval of 1, following 10,000 burn-in iterations and 40,000 sampling iterations. Model convergence was assessed using convergence diagnostic plots. Good convergence was indicated by the stabilization of these plots and a convergence diagnostic value of < 1.01. Comparisons between interventions were visualized using network diagrams. If a closed loop was present, the node-splitting method was adopted to test the inconsistency. Heterogeneity among studies was presented using forest plots. For potential sources of heterogeneity, such as intervention duration, participant training status, and exercise type, subgroup analyses were conducted. Significant findings regarding the comparisons between interventions were presented in league tables, displaying the mean difference (MD) and its corresponding 95% credible interval (CrI). The relative efficacy of interventions was assessed by ranking them based on SUCRA values. A higher SUCRA value represented a greater likelihood of the intervention being more effective than other interventions. Results Study selection results The initial search yielded 11,522 records in total, including 1957 from the Cochrane Library, 2366 from Web of Science, 1721 from PubMed, 2185 from Scopus, 2210 from Embase, and 1083 from EBSCO. After removing 6173 duplicate records, the remaining 5349 records underwent title and abstract screening, during which 5267 were excluded. Subsequently, full texts were successfully retrieved for all 82 remaining articles. Ultimately, these full-text articles were included in the eligibility assessment. After full-text review, 48 articles were excluded, 18 due to ineligible interventions, 28 due to irrelevant outcome measures, and 2 due to insufficient data. Ultimately, 34 studies were included in our review (Fig. 1) 19, 20, 21,25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55.Fig. 1PRISMA literature screening flowchart. PRISMA literature screening flowchart. Basic characteristics of included studies The 34 studies enrolled a total of 1042 participants, comprising 738 men, 184 women, and 120 participants with unreported gender. The sample sizes varied from 12 to 67 participants, and the participants’ ages ranged from 18 to 65 years. Approximately 56% of participants were untrained, while 44% were trained. Among the 42 included BFR intervention groups, 17 groups (40.5%) employed incremental/dynamic pressure, while 25 groups (59.5%) used constant pressure. Pressure was applied during exercise in 26 groups (61.9%), throughout the entire session in 12 groups (28.6%), and during rest in 4 groups (9.5%). In terms of pressure type, relative pressure was used in 29 groups (69.0%), while absolute pressure was used in 13 groups (31.0%). Cuff width was explicitly reported in 25 intervention groups with a range of 5.0–18.0 cm, predominantly 5–8 cm and 8–12 cm. The basic characteristics of the studies are provided in Supplementary Table S3, and BFR intervention strategies are detailed in Supplementary Table S4. Results of the network meta-analysis Convergence diagnostics for all intervention analyses indicated good model convergence (potential scale reduction factor [PSRF] < 1.01) (Supplementary Figs. S1–S11). Furthermore, the inconsistency test using the node-splitting method showed that the consistency assumption was met in this NMA (Supplementary Figs. S12–S18). The overall heterogeneity in this study was low (Supplementary Figs. S19–S53). For subgroup analyses of trained individuals, low-to-moderate intensity training, walking, and cycling, inconsistency tests were not performed because the network structure lacked consistent closed loops that could be utilized to compare direct and indirect evidence. The SUCRA heatmap of different exercise interventions across subgroups in Fig. 2 showed that HIIT + BFR, LICT + BFR, and MICT all ranked high, suggesting that these intervention strategies may yield superior effects.Fig. 2Heatmap. Notes to Fig. 2: The value in each cell represents the SUCRA value (%) for that intervention in the corresponding subgroup. The “R” in parentheses indicates the rank, where R1 denotes 1st place, R2 denotes 2nd place, and so on. A higher SUCRA value represents a greater probability of the intervention ranking high within that subgroup. Colors ranging from blue to red indicate SUCRA values from low to high. “NA” indicates no available data or impossibility to rank within that subgroup. Heatmap. Notes to Fig. 2: The value in each cell represents the SUCRA value (%) for that intervention in the corresponding subgroup. The “R” in parentheses indicates the rank, where R1 denotes 1st place, R2 denotes 2nd place, and so on. A higher SUCRA value represents a greater probability of the intervention ranking high within that subgroup. Colors ranging from blue to red indicate SUCRA values from low to high. “NA” indicates no available data or impossibility to rank within that subgroup. The efficacy of exercise interventions in enhancing VO2max over a 4–6-week intervention period was reported in 18 studies, involving 499 participants (Fig. 3a). Compared with the control, none of the interventions showed statistically significant differences. However, in pairwise comparisons between interventions, HIIT + BFR was superior to HIIT (MD = 1.87, 95% CrI: 0.63–3.10) and LIIT (MD = 6.00, 95% CrI: 2.26–9.68). HIIT was superior to LIIT (MD = 4.12, 95% CrI: 0.61–7.62). LIIT + BFR was superior to LIIT (MD = 2.81, 95% CrI: 1.64–3.96). MIIT + BFR was superior to LIIT (MD = 6.04, 95% CrI: 0.66–11.37) (Supplementary Table S5).Fig. 3(a) Network evidence plot of different blood flow restriction training forms for enhancing maximal oxygen uptake (subgroup of 4–6-week interventions); (b) Network evidence plot of different blood flow restriction training forms for enhancing maximal oxygen uptake (subgroup of 8–12-week interventions); (c) Network evidence plot of different blood flow restriction training forms for enhancing maximal oxygen uptake (subgroup of trained participants); (d) Network evidence plot of different blood flow restriction training forms for enhancing maximal oxygen uptake (subgroup of untrained participants); (e) Network evidence plot of different blood flow restriction training forms for enhancing maximal oxygen uptake (subgroup of low-to-moderate intensity training); (f) Network evidence plot of different blood flow restriction training forms for enhancing maximal oxygen uptake (subgroup of male participants); (g) Network evidence plot of different blood flow restriction training forms for enhancing maximal oxygen uptake (subgroup of mixed-gender participants); (h) Network evidence plot of different blood flow restriction training forms for enhancing maximal oxygen uptake (subgroup of walking interventions); (i) Network evidence plot of different blood flow restriction training forms for enhancing maximal oxygen uptake (subgroup of running interventions); (j) Network evidence plot of different blood flow restriction training forms for enhancing maximal oxygen uptake (subgroup of cycling interventions). (a) Network evidence plot of different blood flow restriction training forms for enhancing maximal oxygen uptake (subgroup of 4–6-week interventions); (b) Network evidence plot of different blood flow restriction training forms for enhancing maximal oxygen uptake (subgroup of 8–12-week interventions); (c) Network evidence plot of different blood flow restriction training forms for enhancing maximal oxygen uptake (subgroup of trained participants); (d) Network evidence plot of different blood flow restriction training forms for enhancing maximal oxygen uptake (subgroup of untrained participants); (e) Network evidence plot of different blood flow restriction training forms for enhancing maximal oxygen uptake (subgroup of low-to-moderate intensity training); (f) Network evidence plot of different blood flow restriction training forms for enhancing maximal oxygen uptake (subgroup of male participants); (g) Network evidence plot of different blood flow restriction training forms for enhancing maximal oxygen uptake (subgroup of mixed-gender participants); (h) Network evidence plot of different blood flow restriction training forms for enhancing maximal oxygen uptake (subgroup of walking interventions); (i) Network evidence plot of different blood flow restriction training forms for enhancing maximal oxygen uptake (subgroup of running interventions); (j) Network evidence plot of different blood flow restriction training forms for enhancing maximal oxygen uptake (subgroup of cycling interventions). The efficacy of exercise interventions in enhancing VO2max over an 8–12-week intervention period was reported in 6 studies, involving 228 participants (35 in control groups) and 8 interventions (Fig. 3b). According to the results in the league table, interventions that significantly outperformed the control in increasing VO2max included MICT (MD = 6.02, 95% CrI: 2.24–9.77), HIIT (MD = 4.49, 95% CrI: 1.39–7.57), and HIIT + BFR (MD = 4.75, 95% CrI: 1.63–7.83). LICT + BFR (MD = 6.09, 95% CrI: −0.16–12.33), HICT + LIRT-BFR (MD = 2.54, 95% CrI: −2.04 to 7.07), and HICT + MIRT (MD = 2.46, 95% CrI: −2.91 to 7.75) did not show statistically significant differences compared with the control (Supplementary Table S6). The efficacy of exercise interventions in enhancing VO2max in trained individuals was reported in 12 studies, involving 327 participants and 7 interventions (Fig. 3c). Since no closed loop existed in the network, no inconsistency test was performed. The analysis results indicated that none of the interventions demonstrated statistically significant differences compared with the control. However, in pairwise comparisons between interventions, HIIT + BFR was superior to HIIT (MD = 1.97, 95% CrI: 0.81–3.10) (Supplementary Table S7). The efficacy of exercise interventions in enhancing VO2max in untrained individuals was reported in 14 studies, involving 465 participants and 9 interventions (Fig. 3d). According to the results in the league table, interventions that significantly outperformed the control in increasing VO2max included LICT + BFR (MD = 5.87, 95% CrI: 0.68–10.96), HIIT + BFR (MD = 5.40, 95% CrI: 2.53–8.28), and HIIT (MD = 4.24, 95% CrI: 1.31–7.17). MICT exhibited no statistically significant difference compared with the control (MD = 5.46, 95% CrI: −0.38–11.28). In pairwise comparisons between interventions, HIIT was superior to LIIT (MD = 4.24, 95% CrI: 0.72–7.74). LIIT + BFR outperformed LIIT (MD = 3.39, 95% CrI: 2.46–4.31). HIIT + BFR was superior to LIIT (MD = 5.4, 95% CrI: 1.47–9.33). LICT + BFR was superior to LIIT (MD = 5.87, 95% CrI: 1.06–10.47). No statistically significant difference between HIIT + BFR and HIIT was noted (Supplementary Table S8). The efficacy of low-to-moderate intensity exercise interventions in enhancing VO2max was reported in 13 studies, involving 344 participants and 5 interventions. The network diagram of the interventions is displayed in Fig. 3e. According to the results in the league table, LICT + BFR was superior to LIIT (MD = 6.29, 95% CrI: 0.54–11.98). LICT + BFR was superior to LICT (MD = 2.56, 95% CrI: 0.06–5.08). LIIT + BFR outperformed LIIT (MD = 3.72, 95% CrI: 2.96–4.48). No statistically significant differences were observed in comparisons between the remaining interventions (Supplementary Table S9). The efficacy of exercise interventions in enhancing VO2max in male participants was reported in 11 studies, involving 295 participants and 6 interventions. The network diagram of the interventions is displayed in Fig. 3f. According to the results in the league table, HIIT + BFR (MD = 3.66, 95% CrI: −1.71 to 9.09), HIIT (MD = 1.97, 95% CrI: −3.41 to 7.41), MIIT + BFR (MD = 3.95, 95% CrI: −1.14 to 9.09), MIIT (MD = 2.80, 95% CrI: −1.93 to 7.59), and MICT (MD = 4.50, 95% CrI: −1.78 to 10.85) exhibited no statistically significant differences compared with the control. In pairwise comparisons between interventions, HIIT + BFR was superior to HIIT (MD = 1.69, 95% CrI: 0.49–2.89), while no statistically significant differences were observed among the remaining interventions (Supplementary Table S10). The efficacy of exercise interventions in enhancing VO2max in participants of mixed genders was reported in 9 studies, involving 269 participants and 9 interventions (Fig. 3g). According to the results in the league table, none of the interventions showed statistically significant differences compared with the control. In pairwise comparisons between interventions, HIIT significantly outperformed LIIT (MD = 4.27, 95% CrI: 0.39–8.13), LIIT + BFR significantly outperformed LIIT (MD = 3.59, 95% CrI: 0.02–7.14), and no statistically significant differences were identified in the comparisons between the remaining interventions (Supplementary Table S11). The efficacy of walking interventions in enhancing VO2max was reported in 7 studies, involving 214 participants and 4 interventions (Fig. 3h). According to the results in the league table, LIIT + BFR outperformed LIIT (MD = 3.59, 95% CrI: 2.66–4.52). LICT + BFR was superior to LIIT (MD = 6.46, 95% CrI: 0.75–12.16) and LICT (MD = 3.09, 95% CrI: 0.71–5.45). No statistically significant differences were observed in LICT + BFR versus LIIT + BFR (MD = 2.86, 95% CrI: −2.74 to 8.51), LICT versus LIIT + BFR (MD = 0.22, 95% CrI: −5.30 to 5.70), and LICT versus LIIT (MD = 3.37, 95% CrI: −2.19 to 8.95) (Supplementary Table S12). The efficacy of running interventions in enhancing VO2max was reported in 6 studies, involving 191 participants and 7 interventions. The network diagram of the interventions is displayed in Fig. 3i. According to the results in the league table, only MIIT + BFR showed a statistically significant advantage over the control (MD = 4.99, 95% CrI: 0.17–9.82), while HIIT + BFR, HIIT, MIIT, LICT + BFR, and MICT showed no statistically significant differences compared with the control. Pairwise comparisons between interventions showed that HIIT + BFR was superior to HIIT (MD = 1.84, 95% CrI: 0.30–3.38), while no statistically significant differences were observed in comparisons between the remaining interventions (Supplementary Table S13). The efficacy of cycling interventions in enhancing VO2max was reported in 8 studies, involving 249 participants and 6 interventions. The network diagram of the interventions is displayed in Fig. 3j. According to the results in the league table, HIIT + BFR, HIIT, and MICT all showed a trend toward improvement compared with the control. Among these, HIIT + BFR significantly enhanced VO2max compared with the control (MD = 5.31, 95% CrI: 2.33–8.31). HIIT (MD = 3.91, 95% CrI: 0.80–7.01) and MICT (MD = 6.26, 95% CrI: 1.82–10.73) also demonstrated a statistically significant advantage over the control. LIIT + BFR and LIIT showed no statistically significant differences compared with the control. Pairwise comparisons between interventions showed that HIIT + BFR was superior to LIIT (MD = 5.30, 95% CrI: 0.75–9.88), MICT was superior to LIIT (MD = 6.27, 95% CrI: 0.76–11.79), and no statistically significant differences were observed among the remaining interventions (Supplementary Table S14). Risk of bias assessment results The risk of bias assessment conducted using ROB 2 showed that, among the 34 studies, 19 (55.9%) were at low risk, 5 (14.7%) raised some concerns, and 10 (29.4%) were at high risk. The risk of bias in the randomization process, missing outcome data, and outcome measurement was generally low, with 91.2%, 91.2%, and 91.2% of studies classified as low risk, respectively. In terms of deviations from intended interventions, 23.5% of studies were rated as high risk. Regarding the selective reporting of results, 26.5% of studies were rated as having some concerns, while 11.8% were rated as high risk. Overall, the methodological quality of the studies included in this review was acceptable. The risk of bias assessment results are presented in the risk of bias assessment plot (Fig. 4) and the risk of bias summary plot (Fig. 5).Fig. 4Risk of bias assessment plot.Fig. 5Risk of bias summary plot. Risk of bias assessment plot. Risk of bias summary plot. Discussion This study employed a Bayesian NMA approach to compare the efficacy of different protocols combining BFR with AE, as well as other training protocols, on VO2max and VO2peak in adults. Overall, HIIT + BFR, LICT + BFR, and MICT ranked highest. Specifically, HIIT + BFR showed favorable relative rankings in the subgroups of short-term interventions, running-related interventions, and trained individuals. LICT + BFR showed better performance in subgroups of untrained individuals, walking interventions, and low-to-moderate intensity exercises. MICT exhibited a more consistent improvement in subgroups of 8–12-week interventions, cycling interventions, and in well-trained individuals. In contrast, LIIT alone did not demonstrate significant advantages in multiple subgroups. These subgroup findings reflect average group-level differences across study-defined contexts and should not be interpreted as evidence that a particular protocol is optimal for a specific individual. A key finding of this study is that VO2max improvement does not rely entirely on high mechanical load. Traditional views generally hold that higher intensity or greater training volume are the primary prerequisites for improving aerobic capacity. However, this study suggests that, in specific populations and for specific forms of exercise, BFR enables training with lower external loads to achieve effects that are close to or even partially comparable to those of higher-intensity training. This finding aligns with the overall direction of previous research. Existing research has demonstrated that low-intensity AE plus BFR can enhance VO2max, extend time to exhaustion, and improve certain aspects of endurance performance. Relevant systematic reviews have also suggested that BFR combined with AE may offer additional benefits compared with similar exercises without BFR 21,26,56, 57, 58, 59. Therefore, from a training-practice perspective, BFR should not be viewed as a replacement for traditional endurance training, but rather as a potential supplementary strategy for improving CRF in specific training contexts. Among various BFR protocols, LICT + BFR showed superior performance in the subgroups of untrained individuals, walking interventions, and low-to-moderate intensity exercise. This finding holds strong practical implications. Untrained individuals typically possess greater training potential. Their lower baseline CRF means that even relatively mild exercise may yield significant adaptive improvements. Meanwhile, BFR increases local hypoxia and metabolite accumulation during low-load exercise, elevating heart rate, cardiac output, oxygen consumption during recovery, and lactate response. This amplifies the training stimulus without significantly increasing mechanical stress.60,61 For individuals with poor baseline fitness, joint sensitivity, those in rehabilitation, or those unable to tolerate high-intensity training, LICT + BFR may hold significant application potential. In contrast, HIIT + BFR showed favorable relative rankings in short-term interventions, running-related interventions, and among trained individuals. In the 4–6-week subgroup, HIIT + BFR had the highest SUCRA value, indicating the most favorable relative ranking among the compared interventions, and pairwise comparisons showed that it was significantly superior to HIIT and LIIT. However, none of the interventions differed significantly from the no-exercise control. This lack of statistical significance may partly reflect the wider credible intervals resulting from the between-study heterogeneity and uncertainty in indirect evidence within the Bayesian random-effects NMA, particularly given the sparse network, limited number of studies, and small sample sizes. Thus, these findings only suggest that HIIT + BFR may have potential as a short-term training strategy. In the running-related subgroup, HIIT + BFR was superior to HIIT but did not show advantages across all comparisons, suggesting that its effects may be influenced by sample size, training background, and training dose. Although previous studies have suggested that HIIT + BFR may improve maximal aerobic power, sprint capacity, and certain aspects of endurance performance over the short term,30,38,49 its short-term effects should be interpreted cautiously in light of the present findings. This study also showed that MICT performed more consistently in 8–12-week interventions, cycling interventions, male participants, and some well-trained individuals. This suggests that as the training cycle extends, traditional MICT still holds irreplaceable value. The improvement in VO2max depends not only on local metabolic stimulation but also on integrated central-peripheral adaptations, such as increased cardiac output, improved peripheral vascular function, increased capillary density, and enhanced oxygen extraction capacity of skeletal muscle. These changes typically require sufficient training volume and duration 62, 63, 64. Therefore, in long-term interventions, MICT can continuously promote aerobic adaptations by virtue of its high completion rate, stable stimulation pattern, and good tolerability. Particularly among well-trained individuals or individuals with a clear sport-specific training background, the subjects’ high baseline adaptation levels and relatively limited room for further improvement may diminish the additional benefits provided by BFR. In such cases, standardized and sustainable traditional endurance training may still hold greater practical value. Previous studies on trained individuals have also suggested that BFR is not necessarily superior to the same training without BFR. Its effectiveness must be evaluated taking into account the training context, sport-specific requirements, and program scheduling.65,66 In contrast, LIIT consistently ranked low across multiple subgroups, suggesting that LIIT alone may be insufficient to raise VO2max. This may be because, without sufficient total training volume, adequately high relative intensity, or the metabolic amplification effect provided by BFR, it is difficult for LIIT to fully activate the central and peripheral adaptation mechanisms closely linked to improvements in CRF. Previous studies have similarly shown that LIIT without BFR often yields only limited improvements in VO2max.11,21,67 These findings suggest that when the primary goal is to enhance VO2max, simply reducing training intensity and adopting an interval training plan does not necessarily yield desirable benefits. The effectiveness of the training protocol remains dependent on the comprehensive balance between training intensity and volume and physiological load. From a physiological perspective, BFR may primarily promote aerobic adaptation by enhancing local hypoxia, increasing metabolic stress, and amplifying training signals. Acute BFR interventions can increase hypoxia and metabolite accumulation during exercise and are associated with higher heart rate, cardiac output, VO2 during recovery, and lactate responses.61 Long-term interventions may promote mitochondrial biogenesis and angiogenesis and improve endothelial function by upregulating relevant signaling pathways, including PGC-1α, VEGF, hypoxia-inducible factor 1-alpha (HIF-1α), VEGF receptor 2 (VEGFR-2), and endothelial nitric oxide synthase (eNOS).68 Furthermore, BFR may further enhance aerobic metabolic adaptations in skeletal muscle by improving glucose transport, local oxygen utilization efficiency, and antioxidant capacity.69 Collectively, these mechanisms suggest that BFR enables low-to moderate-intensity exercise to generate higher levels of metabolic stimulation by altering the local physiological environment during training. However, this study has several limitations that must be carefully considered when interpreting the results. First, the overall sample size of the included studies is small, with some subgroups comprising only a few studies, leading to considerable uncertainty in effect estimates. Second, although the ranking results from the NMA may serve as a reference for comparing different protocols, the SUCRA values essentially reflect relative ranking probabilities and cannot replace effect sizes and their credible intervals. Therefore, the determination of the optimal protocol should not be based solely on the ranking results. Third, the included studies exhibit significant variations in BFR pressure settings, cuff width, timing of pressure application, exercise intensity, and training duration, suggesting that clinical heterogeneity may still exist within the network nodes. Fourth, due to the absence of closed loops in some subgroups, it is impossible to conduct inconsistency tests, which, to some extent, limits the robustness of the indirect comparison results. Fifth, whether VO2max/VO2peak was reassessed during the intervention was not consistently reported across the included studies; therefore, potential learning, familiarization, or pacing effects from repeated testing could not be evaluated. Sixth, existing studies remain insufficient regarding the reporting of safety, adherence, and adverse events, which are precisely the key prerequisites for the widespread application of BFR across different populations. Therefore, a more appropriate formulation at this stage is that BFR combined with AE has the potential to enhance VO2max in specific contexts, rather than that a single, universal, and directly applicable optimal prescription has been established. Finally, the present NMA compared the average effects of different training protocols at the group level and therefore could not determine whether individuals differed in their true responses to training. In principle, heterogeneity in response among individuals can be examined in randomized trials with a time-matched no-exercise control group by comparing change-score variability between the intervention and control groups.70 This approach has also been used in previous meta-analyses 71, 72, 73. However, it relies on several assumptions and should be interpreted cautiously. The N-of-1 design may provide more direct evidence, but its use in exercise training is limited by carryover effects, detraining effects, and the feasibility of repeated intervention periods.74 These analyses were not performed in the present study; therefore, our findings should be interpreted as group-level comparative effects. In conclusion, at the group level, BFR combined with AE may improve VO2max/VO2peak in some intervention settings, while MICT remains an effective conventional training strategy. HIIT + BFR was associated with more favorable average effects in short-term and running-based subgroups, whereas LICT + BFR showed higher average rankings in studies involving untrained participants, walking-based exercise, and low-to moderate-intensity training. MICT was associated with more favorable average effects in longer-duration and cycling-based interventions. These subgroup findings should not be interpreted as evidence for selecting an optimal protocol for a specific individual. Further high-quality studies are needed to inform protocol-specific recommendations. Perspective Previous studies and reviews have suggested that AE combined with BFR may improve aerobic capacity, but most evidence has focused on pairwise comparisons and has not clarified the relative group-level effects of different protocols across study contexts 14,56, 57, 58, 59,65,66. By integrating 34 RCTs involving 1042 participants, the present NMA extends this literature by comparing BFR-based aerobic protocols with conventional exercise strategies within a unified framework. The findings suggested that the relative average effects of the interventions were not uniform across study-level subgroups. HIIT + BFR was associated with higher average rankings in studies involving shorter intervention durations, trained participants, and running-based exercise, whereas LICT + BFR was associated with higher average rankings in studies involving untrained participants, walking-based exercise, and low-to moderate-intensity training. MICT was associated with more favorable average effects in studies with longer intervention durations and cycling-based exercise. These subgroup-level patterns reflect aggregate study-level associations and should not be interpreted as evidence that participant characteristics can be used to select an optimal protocol for a particular individual. From a sports medicine perspective, BFR should therefore be viewed as a potential complement to, rather than a replacement for, conventional endurance training. Future trials should standardize BFR pressure, cuff characteristics, safety reporting, and adherence monitoring to support more robust protocol-specific recommendations for clinical and athletic settings. Prospero registration number CRD420261278478. Data availability statement All data generated or analysed during this study are included in this published article and its supplementary information files. Ethics approval statement Not applicable. Author contributions Hangxin Chen: Conceptualization, Methodology, Software, Data curation, Writing- Original draft preparation. Feng Xu: Visualization, Investigation. Ziqi Fan: Software, Validation. Yangyang Liu: Software, Validation. Dongxu Gao: Supervision, Writing- Reviewing and Editing, and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript. Funding statement The authors declare that no funds, grants, or other support were received during the preparation of this manuscript. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.