This systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines 20 by two of the authors and was prospectively registered in the International Prospective Register of Systematic Reviews (PROSPERO; registration number: #420251132262). A comprehensive search of EMBASE, MEDLINE/PubMed, and Cochrane CENTRAL was performed from database inception to July 2025. In addition, manual searches of the reference lists of relevant reviews and included studies were conducted to identify any additional eligible articles. Relevant Medical Subject Headings (MeSH) terms and free-text keywords were applied to identify studies related to participants (e.g., hypertension, pre-hypertension, elevated blood pressure), interventions (e.g., exercise, training), and outcomes (e.g., blood pressure, hypertension), as outlined in Table 1.

Study selection followed the PICOS framework (Population, Intervention, Comparison, Outcome, and Study design). (1) Population: adults aged ≥18 years diagnosed with hypertension or prehypertension, including individuals with metabolic syndrome who met the BP criteria, were eligible. Prehypertension was defined as SBP ≥120 mmHg and/or DBP ≥80 mmHg, whereas hypertension was defined as SBP ≥140 mmHg and/or DBP ≥90 mmHg. Studies involving pregnant women or individuals with stroke, heart failure, or chronic kidney disease were excluded. (2) Intervention: eligible studies included at least one supervised (SUP) training group and one non-SUP training group. SUP was defined as having ≥ two-thirds (≥0.67) of total training sessions conducted under supervision (e.g., guided by an instructor), whereas non-SUP was defined as having ≤ one-third (≤0.33) of sessions conducted under supervision. Studies were excluded if the supervision ratio fell between 0.33 and 0.67 or was not clearly defined 21 to minimize overlap between groups and improve conceptual separation between predominantly supervised and predominantly self-directed exercise interventions. (3) Comparison: the comparator was a group receiving non-supervised exercise training. Studies that mentioned only “standard care” without specific exercise recommendations were excluded. (4) Outcomes: studies were required to report both baseline and post-intervention BP measurements for SBP and DBP. (5) Study design: only randomized controlled trials (RCTs) were included. Systematic reviews, meta-analyses, conference abstracts, case reports, dissertations, observational studies, and studies not published in English were excluded.

Two reviewers (*** and **) independently screened the titles and abstracts of all retrieved records to identify studies meeting the inclusion criteria. Full-text articles were obtained for studies deemed potentially eligible by two reviewers (*** and **) independently. All references were imported into EndNote 21 (Clarivate, Philadelphia, PA, USA), and duplicate entries were removed. Discrepancies between reviewers were resolved through consensus, with arbitration by a third reviewer when necessary. Data extraction was conducted by one reviewer using a form developed for this purpose. The extracted variables included study identifiers (authors and year), group allocation, inclusion criteria, participant characteristics, medication use, BP type and measurement methods, intervention characteristics, training content, exercise frequency and duration, follow-up duration, and quantitative outcome data (means, SDs, and sample size) used for the meta-analysis.

The methodological quality of the included studies was assessed using the Physiotherapy Evidence Database (PEDro) scale. 22 This tool comprises 11 items, including eligibility criteria, random allocation, concealed allocation, baseline comparability, blinding of subjects, blinding of therapists, blinding of assessors, adequate follow-up, intention-to-treat analysis, between-group comparisons, and reporting of point estimates with measures of variability. The final 10 items contribute to the total PEDro score. Based on these scores, study quality was classified as excellent (9-10), good (6-8), fair (4-5), or poor (≤3). When available, PEDro scores were extracted directly from the PEDro database (https://pedro.org.au/english/resources/pedro-scale/), which included committee-assigned scores for three of the included studies .12,17,23 For studies not listed in the database, two reviewers independently assessed quality in accordance with the PEDro guidelines. 22 Discrepancies were resolved through discussion, with a third reviewer serving as arbiter when consensus could not be reached.

The means and SDs of the reported outcomes were extracted by the first author into a Microsoft Excel 2019 spreadsheet and independently verified by the second author. Discrepancies were resolved through discussion with the third author. When data were available only in graphical format and could not be obtained from the study authors, mean and SD values were extracted using WebPlotDigitizer. 24 Data synthesis and analysis were performed using Review Manager (RevMan) version 5.4 (Nordic Cochrane Center, Cochrane Collaboration, Copenhagen, Denmark). For SBP and DBP, the mean differences and corresponding SDs of the mean differences were calculated. The SDs of the mean differences were derived using published equations, 20 applying a conservative correlation coefficient (r = 0.5).

A random-effects model was applied to all analyses. Statistical heterogeneity was quantified using the inconsistency index (I²), with values >50% indicating substantial heterogeneity and values <50% indicating low heterogeneity. Statistical significance was set at p < 0.05. In addition, subgroup analyses were performed according to exercise programs, including continuous aerobic exercise training, high-intensity interval training (HIIT), isometric handgrip training, and combined training, to further explore potential sources of heterogeneity and differences in intervention effects. Pooled effect estimates for SBP and DBP between comparison groups were presented as forest plots with 95% confidence intervals (CIs). Publication bias assessment using funnel plots was planned only when sufficient studies were available, as interpretation of funnel plots may be unreliable with a small number of studies.

The studies by Gorostegi-Anduaga et al. 17 and Molmen-Hansen et al. 25 included three and two SUP groups, respectively. We divided the non-SUP group into subgroups for comparison in the meta-analysis. Specifically, the non-SUP group in Gorostegi-Anduaga et al. comprised 40 participants, whom we divided into three subgroups (n = 13, 13, and 14). In Molmen-Hansen et al., the non-SUP group comprised 25 participants, whom we divided into two subgroups (n = 13 and n = 12).

A total of 8,520 reports were retrieved: 8,519 from the three databases and 1 via manual searching. After removing 1,649 duplicate reports, the remaining 6,871 reports were screened. Fifty-five reports matched the inclusion criteria for full-text download; however, two could not be retrieved. Of the 53 full-text articles successfully downloaded, 41 reports were subsequently excluded for reasons such as the wrong comparator, the wrong study design, no exercise intervention, the wrong population, no outcome of interest, and the wrong publication type. Appendix 1 details the excluded reports and the reasons. Consequently, 12 reports were identified; however, only 11 unique studies were included in the systematic review because one trial registry record 26 and one journal article 13 referred to the same study (Fig. 1). Ten studies 12,13,17,18,23,25,27-30 were included in the final quantitative synthesis because one study 31 did not report SD values. The obtained studies were published between 2003 and 2024.

Fig. 1.

Flow diagram showing search and selection of articles for systematic review.

The average PEDro score across the 11 studies was 5.3 ± 1.6, with a maximum of 7 and a minimum of 1 (Table 2). Thus, the reviewed studies have methodological quality ranging from fair to good. Due to the limited number of included studies in most subgroup analyses, publication bias was not formally assessed.

The youngest lower limit reported was 18 years, and the oldest upper limit was 70 years (one study included ≥60 years). Across the included studies, the participants were predominantly middle-aged to older adults with overweight to obese body mass index (BMI) values. Five studies 12,17,18,25,31 specifically included participants with stage 1-2 hypertension, while six studies 13,23,27-30 did not report the hypertension stage in detail. Notably, one study 30 recruited participants with metabolic syndrome, of whom 91.3% had hypertension.

In the 11 included studies, a total of 462 participants were initially allocated to the SUP group, with a mean of 42.0 ± 36.7 subjects per study. After accounting for dropouts, the final sample size of the SUP group was 396, corresponding to an average of 36.0 ± 36.4 participants per study. For the non-supervised (non-SUP) group, 293 participants were originally allocated, with a mean of 26.6 ± 11.8 per study. After dropouts, the final number of participants was 252, with an average of 22.9 ± 12.1 per study.

For the 10 studies included in the meta-analysis, a total of 378 participants were allocated to the SUP group. The mean number of participants per study was 37.6 ± 28.4 for continuous aerobic exercise training, 31.8 ± 11.5 for HIIT, 9.0 ± 1.4 for isometric handgrip training, and 22.5 ± 9.2 for combined training. In comparison, 229 participants were allocated to the non-SUP group, with corresponding means of 22.2 ± 15.2 for continuous aerobic exercise training, 14.8 ± 2.9 for HIIT, 7.5 ± 2.1 for isometric handgrip training, and 22.0 ± 9.9 for combined training.

Regarding the training interventions, 6 studies 12,17,25,28,30,31 with fair quality were considered continuous aerobic exercise training, 3 studies 17,23,25 with fair quality used HIIT, 2 studies 13,18 with good quality used isometric handgrip, and 2 studies 27,29 with good quality performed combined training (resistance training, aerobic, and stretching exercises). Regarding the outcome measures, six studies 12,17,18,27,30,31 assessed resting BP only, one study 25 reported 24-h BP, one study 29 assessed both resting and central BP, one study 13 reported resting, central, and 24-h BP, and another study 23 assessed resting, 24-h, and exercise-induced BP. In addition, one study 28 reported resting, 24-h, daytime, and nighttime BP. Table 3 summarizes the interventions across the 11 included studies. The non-SUP groups varied substantially across the 11 included studies, with some performing exercise programs similar to those of the SUP groups in a home-based or partially supervised setting, 12,13,18,27,30-32 while others received only general physical activity recommendations, health education, or standard care rather than equivalent exercise interventions, 17,23,25,28

The effects of SUP versus non-SUP exercise training on SBP and DBP were synthesized.

Ten studies 12,13,17,18,23,25,27-30 were scrutinized to examine the effects of SUP training on SBP. The pooled results showed significant differences (p = 0.006, I2 = 74%) between the SUP and non-SUP groups (Fig. 2). The mean difference favored the SUP group (MD = -3.46 mmHg; 95% CI: -5.94, -0.97). The subgroup analysis results indicated that continuous aerobic exercise training (p = 0.003, I2 = 64%) favors SUP in reducing SBP. However, no significant differences were observed for HIIT (p = 0.63, I2 = 60%), isometric handgrip (p = 0.38, I2 = 93%), or combined training (p = 0.60, I2 = 26%).

Fig. 2.

Forest plots of systolic blood pressure reductions.

Ten studies 12,13,17,18,23,25,27-30 were similarly examined to determine whether SUP training had any effect on DBP compared with non-SUP training. The pooled results showed no significant differences (p = 0.41, I2 = 89%) between the SUP and non-SUP groups (Fig. 3). The mean difference was -0.99 mmHg (95% CI: -3.38, 1.39). The results of subgroup analyses indicated that continuous aerobic exercise training (p = 0.15, I2 = 80%), HIIT (p = 0.88, I2 = 36%), isometric handgrip (p = 0.90, I2 = 96%), and combined training (p = 0.81, I2 = 78%) also had a similar effect on the change of DBP.

Fig. 3.

Forest plots of diastolic blood pressure reductions.

The findings demonstrated that SUP interventions were associated with greater reductions in SBP than non-SUP interventions (mean difference = -3.46 mmHg; 95% CI: -5.94, -0.97; p = 0.006). SUP interventions may be more effective than non-SUP training because SUP enhances adherence, ensures quality training, and provides motivation and social support. 33,34 With respect to adherence, patients in non-SUP settings may either fail to perform the prescribed exercises correctly or neglect them altogether, whereas SUP generally ensures stricter adherence to the exercise program, reducing the risk of missed sessions or insufficient training. 35,36 In addition, SUP enables patients to exercise with proper posture and intensity, thereby preventing ineffective training or injuries caused by improper technique. 35 Furthermore, a supervised environment often offers greater motivation, encouragement, and social support, which may facilitate long-term exercise participation. From a clinical perspective, the findings suggest that patients with elevated SBP may respond more favorably to supervised exercise intervention.

The current meta-analysis showed no clear differences between SUP and non-SUP interventions in reducing DBP (mean difference = -0.99 mmHg; 95% CI: -3.58, 1.39; p = 0.41), with high heterogeneity. This could indicate that both approaches are equally effective in lowering DBP. Since DBP is primarily influenced by peripheral arteriolar resistance, which adapts more slowly, short-term interventions, even with supervision, may not be sufficient to produce measurable changes. Alternatively, the absence of a difference may be due to insufficient statistical power or high heterogeneity, which limit the ability to detect between-group effects. In this study, heterogeneity in DBP outcomes reached as high as 89%, suggesting that the lack of a difference is more likely attributable to insufficient evidence rather than to truly equivalent effects. Among the included studies, 7 of 10 involved participants on antihypertensive medication, of which 2 reported 100% medication use. Only one study completely excluded pharmacological treatment, thereby providing the clearest evidence of exercise's isolated effect. In contrast, 2 studies did not report medication status, which limits the transparency and applicability of their findings. The widespread use of antihypertensive therapy among participants suggests that BP, particularly DBP, was already well controlled at baseline. Under such conditions, the additional margin for BP reduction through exercise, whether supervised or not, may have been limited, which could explain the similarity of effects between SUP and non-SUP interventions. Thus, medication use represents an important source of heterogeneity in exercise trials for hypertension. Future studies should more rigorously control for or stratify analyses by medication status to better isolate the independent effects of exercise.

Two main reasons may explain why differences were observed for SBP but not for DBP. First, regulation of SBP is largely influenced by arterial elasticity and cardiac output; therefore, SBP may be inherently more sensitive to general lifestyle modifications, including exercise and dietary interventions .9 Given this characteristic, improvements in SBP may be achieved more in SUP settings, which could partly explain why the additional benefit of supervision appears pronounced for SBP. In contrast, DBP is primarily determined by the peripheral resistance of arterioles, a process that requires a longer period for vascular remodeling and adaptation of endothelial function. 9 As a result, the magnitude of DBP reduction is typically smaller and less pronounced than that observed for SBP. Second, previous meta-analyses have shown that exercise interventions reduce SBP by approximately 6–8 mmHg, whereas DBP decreases by only 3–5 mmHg on average. 37,38 This general observation indicates that the absolute reduction in SBP is usually greater than that in DBP, making differences between SUP and non-SUP conditions more detectable for SBP than for DBP.

To better understand the influence of supervision across different exercise training, subgroup analyses were conducted. The included training programs encompassed continuous aerobic exercise training, HIIT, isometric handgrip, and combined training. Subgroup analysis revealed that the SUP was associated with greater reductions in SBP than non-SUP interventions, evident only with continuous aerobic exercise training (mean difference = -4.58; 95% CI: -7.62, -1.53; p = 0.003), whereas no significant differences were observed for the other exercise training modalities.

Continuous aerobic exercise training typically emphasizes prolonged sessions to achieve a sufficient total exercise volume. The greater success of SUP continuous aerobic exercise training can be attributed in part to enhanced regulation of exercise intensity and improved adherence to prescribed training protocols. Supporting this, network meta-analysis 39 in patients with type 2 diabetes mellitus demonstrated that SUP aerobic exercise was superior to both non-SUP aerobic and non-SUP resistance exercise in reducing glycated hemoglobin (HbA1c) and body weight. Furthermore, in children and adolescents, supervised aerobic activity was shown to achieve significantly greater improvement in cardiorespiratory fitness. 40 Taken together, these findings underscore the importance of professional SUP in continuous aerobic exercise programs, not only to ensure appropriate training intensity but also to enhance adherence, thereby optimizing clinical outcomes in patients with hypertension and related chronic conditions.

The subgroup analysis of HIIT did not reveal consistent differences between SUP and non-SUP interventions. Although some evidence suggested that SUP HIIT may produce greater reductions in BP, this effect was not uniformly observed across trials. The inconsistency might be due to confounding by concomitant dietary intervention and widespread antihypertensive medication use, which likely masked the independent contribution of SUP exercise. Meanwhile, the varying training lengths across studies may have limited the ability to detect significant changes.

These findings suggest that SUP training may be associated with favorable BP outcomes, while also highlighting the complexity of isolating its independent effect. Factors such as exercise intensity, intervention duration, participants’ baseline health status, and concurrent co-interventions (e.g., diet, medication) appear to play critical roles in shaping the BP response. This variability suggests that while supervision may enhance adherence and ensure exercise fidelity, its direct impact on BP outcomes cannot be fully disentangled without controlling for these interacting variables.

In the subgroup analysis of isometric handgrip, only two studies were included, which indeed limits the statistical power of the meta-analysis. A closer examination of these two studies reveals contrasting findings. The non-SUP group in Farah et al. 13 did not show any meaningful reduction in BP because participants may have been unable to perform the handgrip exercise correctly or consistently throughout the intervention period. In contrast, Gordon et al. 18 found that both SUP and non-SUP interventions led to clinically significant reductions in BP. A notable methodological difference between the two studies is that the non-SUP group in Gordon et al.’s trial 18 used instructional videos, potentially enhancing adherence compared with the non-SUP group in Farah et al.’s study, 13 which lacked such guidance.

In the subgroup analysis of combined training, no significant differences were observed between SUP and non-SUP interventions in reducing SBP. This result may be explained by the two included studies not demonstrating notable reductions in BP, thereby limiting the ability to detect additional benefits of supervision. However, previous research has shown that combined training can effectively lower BP. 41 The discrepancy may be attributable to the participant characteristics in these studies, as all individuals were on regular antihypertensive medication and their BP had already been maintained within the normal range, leaving limited room for further reductions. Nevertheless, evidence indicates that combined training is effective in improving other clinically relevant outcomes, including aerobic capacity, muscular strength, functional fitness, and body composition. 38 These adaptations are particularly important in hypertensive populations, as they contribute to improved cardiovascular health and reduced long-term morbidity.

The high heterogeneity suggests that the magnitude of the effect varied substantially across studies, limiting the robustness of this finding. Potential sources of heterogeneity include differences in participant characteristics, exercise protocols, and control conditions. Participants’ mean ages ranged from 35.5 to 61.7 years, and sex distribution varied widely, with some studies including only women (e.g., Bonet et al. 12) and others almost exclusively men (e.g., Islami et al. 27). Conditions of non-SUP were also heterogeneous, with some trials providing only general lifestyle advice, others incorporating dietary modification, and some allowing self-selected or partially supervised exercise. These variations likely influenced the relative differences between SUP and non-SUP groups. Moreover, the interventions differed substantially, including continuous aerobic exercise training, HIIT, isometric handgrip, and combined training, with considerable variation in duration and intensity. Subgroup analysis was performed in the current study to provide clearer insights into the effects of different exercise programs. Nonetheless, the overall trend suggests a possible association between SUP exercise and SBP, and future large-scale, rigorously designed studies are warranted to confirm these findings.

Several limitations should be acknowledged. First, the hypertension stages of participants varied across the included studies, ranging from stage 1 to stage 2 hypertension, to unspecified hypertension stages, with one study 30 including participants diagnosed with metabolic syndrome. Given the limited number of available studies on this topic, aggregating them in a meta-analysis may represent a necessary initial step to explore the effects of supervision among individuals with hypertension. Second, the exercise prescriptions for the non-SUP group were not entirely consistent with those provided to the SUP group. In some studies, non-SUP interventions differed in exercise intensity, structure, or delivery approach and, in some cases, consisted primarily of educational or general physical activity recommendations. Consequently, some observed effects may be attributable to differences in exercise protocols rather than the presence or absence of supervision alone. Moreover, the thresholds of 0.33 and 0.67 were adopted based on previous literature 21 and were intended to provide a pragmatic distinction between predominantly supervised and predominantly self-directed exercise interventions. However, we acknowledge that supervision is a multifaceted construct that may not be fully captured by this dichotomized classification. Finally, three of the eleven included studies were rated as fair methodological quality, and one study was rated as poor methodological quality, which may have affected the robustness and generalizability of the findings. Although the PEDro scale is widely used in physical therapy research, it primarily evaluates methodological quality rather than risk of bias. In exercise intervention studies, blinding of participants and therapists is often not feasible, which may influence PEDro scores and should be considered when interpreting the findings.