In adults with prehypertension or hypertension, every exercise mode lowered resting blood pressure against non-exercise controls, and combined training (−12.05/−6.20 mmHg) and HIIT (−10.97/−6.42) ranked highest while isometric training (−7.45/−4.09) ranked with dynamic resistance training near the bottom, in a Bayesian network meta-analysis of 105 RCTs (6734 participants; 23 trials high risk of bias; GRADE mostly low to moderate); weekly volume showed a shallow U-shaped curve with the largest fall near 830 MET-min a week
Main caveat: The authors': no subgroup analyses by sex, medication use or weekly dose when comparing modes, which may bias the mode comparison.
Xin and colleagues compared seven exercise modes for lowering resting blood pressure in adults whose systolic pressure was at least 120 mmHg or who were treated for hypertension, and modelled how the fall in pressure changes with weekly exercise volume in MET-minutes.
They included 105 RCTs with 6734 participants that compared at least 4 weeks of exercise with a non-exercise or minimal-activity control. Against control, all seven modes lowered systolic and diastolic pressure.
Combined aerobic plus resistance training and HIIT produced the largest falls and ranked highest; yoga and tai chi came next; aerobic, isometric and dynamic resistance training produced smaller falls and ranked lowest.
Volume and effect followed a nonlinear, U-shaped curve: a fall became detectable from 52 MET-min a week, was largest near 830 MET-min a week and flattened at higher volumes, with each mode peaking at a different volume. Larger falls went with higher baseline pressure.
The authors recommend combined training as the first choice for this population and rate most of their evidence low to moderate in certainty.
Participants
105 RCTs, 6734 participants; age range 19.6-76.0 years and BMI 22.6-34.8 kg/m² as reported; aerobic training in 59 trials and combined training in 19; the number of isometric trials is not given in the text.
Inclusion: adults with systolic pressure of 120 mmHg or more (elevated, stage 1 or stage 2 by the 2017 ACC/AHA categories) or diagnosed and medicated hypertension; obesity alone allowed.
Excluded: diabetes, rheumatoid arthritis, cancer, orthopaedic disease; gestational, white-coat, pulmonary and adolescent hypertension; non-English papers; standard-care controls.
Protocol
Search: five databases to April 2025 plus reference lists; trials published January 1990 to April 2025.
Comparison: each mode against non-exercise or minimal-activity controls (light stretching or education), and modes against each other where trials compared them; how much of the network is direct is not stated in the text (network plot only), and GRADE rated indirect comparisons lower.
Outcomes: change in resting systolic and diastolic pressure (pre-post change scores, SD imputed with r = 0.5).
Dose: MET × minutes per session × sessions per week from the Compendium of Physical Activities, grouped at 0, 300, 600, 900, 1200 and 1500 MET-min a week; quadratic model chosen by fit.
Analysis: Bayesian random-effects NMA in MetaInsight (gemtc), SUCRA ranking, node-splitting for inconsistency; MBNMAdose for dose-response; sensitivity analyses excluding high risk of bias, small trials and trials of 8 weeks or less; network meta-regression on duration, sex proportion, age, baseline pressure, publication year, frequency, session length and BMI.
Key Results
| Mode vs control | Systolic, mmHg (95% CrI) | SUCRA systolic | Diastolic, mmHg (95% CrI) | SUCRA diastolic |
|---|---|---|---|---|
| Combined training | −12.05 (−15.08 to −9.05) | 88.10% | −6.20 (−7.79 to −4.62) | 79.76% |
| HIIT | −10.97 (−14.97 to −6.95) | 75.61% | −6.42 (−8.68 to −4.16) | 82.23% |
| Yoga | −10.02 (−13.80 to −6.22) | 65.64% | −5.68 (−7.70 to −3.71) | 67.98% |
| Tai chi | −10.10 (−14.88 to −5.39) | 65.11% | −5.76 (−8.22 to −3.30) | 68.68% |
| Aerobic | −8.19 (−9.97 to −6.44) | 40.56% | −4.57 (−5.49 to −3.64) | 40.13% |
| Isometric | −7.45 (−10.81 to −4.07) | 32.27% | −4.09 (−5.89 to −2.20) | 30.82% |
| Dynamic resistance | −7.46 (−10.73 to −4.20) | 32.71% | −4.10 (−5.89 to −2.28) | 30.39% |
| Weekly volume (all modes) | Systolic, mmHg (95% CrI) | Diastolic, mmHg (95% CrI) |
|---|---|---|
| 52 MET-min/week | first volume with a significant fall | — |
| 600 (WHO lower bound) | −9.26 (−10.99 to −7.51) | −5.21 (−6.14 to −4.24) |
| ≈830 (largest effect) | −10.01 (−11.69 to −8.30) | −5.62 (−6.59 to −4.70) |
| 1200 (WHO upper bound) | −8.27 (−10.92 to −5.54) | −4.62 (−6.18 to −3.14) |
| Beyond (e.g. 1470) | plateau | plateau |
Volume of largest systolic effect by mode: aerobic 720 MET-min/week (−8.91), dynamic resistance 590 (−7.52), combined 1100 (−12.41), HIIT 530 (−11.85), tai chi 580 (−12.01), yoga 790 (−14.79). Isometric: not modelled, because its protocols sat mostly below 300 MET-min a week.
Head-to-head contrasts (the league table) are in a figure not reproduced in the text; the text says aerobic training 'outperformed both RT and IET' and that isometric training was less effective than aerobic in this population, but gives no contrast or interval, so whether that gap is significant is not in the text. By the point estimates, isometric and aerobic training differ by 0.74 mmHg systolic and 0.48 mmHg diastolic (our arithmetic).
Moderators: higher baseline systolic and diastolic pressure went with larger falls; age, duration, sex proportion, year, frequency, session length and BMI did not moderate significantly (coefficients in the supplement).
Sensitivity analyses: comparisons with control stayed significant in sensitivity analyses, but the order of dynamic resistance vs isometric, tai chi vs yoga and aerobic vs yoga changed, and the certainty of those rankings was downgraded. Node-splitting found inconsistency only for aerobic vs yoga (diastolic). The diastolic funnel plot was visually asymmetric.
Practical Application
For someone with high-normal or raised pressure, any regular exercise lowers it, and combined aerobic-plus-strength training or intervals did so most (about −11 to −12 mmHg systolic against no exercise). Across modes the largest fall came near 830 MET-min a week and more did not add; the authors translate the aerobic optimum (720 MET-min) into about 150-163 min a week of moderate aerobic exercise and the combined-training optimum (1100) into about 204-224 min.
Where this differed from [SYNTH-88]: until 2026-11-10 the guide stated that isometric training ranked first of five modes for resting pressure (REF-1150, 270 RCTs of mostly untrained adults with any baseline pressure); it now reports that ranking beside this network's and says the three networks agree that isometric training lowers pressure against control, not that it is the best mode.In this network, restricted to prehypertension and hypertension and with yoga and tai chi added as separate modes, isometric training sat at the bottom for systolic pressure (SUCRA 32.27%, level with dynamic resistance training at 32.71%), below aerobic training. What does not differ: isometric training still lowered pressure against control here, by about the amount REF-1150 gave isometric training overall and handgrip (−7.45/−4.09 here; −8.24/−4.00 and −7.10/−3.46 there); REF-1150's isometric ranking rested on indirect comparisons only, indirect comparisons here carried lower certainty, and this paper does not report whether the isometric-aerobic gap (under 1 mmHg by the point estimates) is significant.So isometric holds remain an evidence-backed add-on; "the best mode for blood pressure" is not supported for people who already have raised pressure. Since 2026-11-10 SYNTH-88 carries this as the base's inference, not tested: with raised pressure the first lever is aerobic training together with strength training, and holds are the add-on for when time is short.
Inference, not tested here: for a masters athlete with raised pressure who already rides or runs, adding two strength sessions (turning aerobic training into combined training) is the change this network ranks highest; handgrip or wall-squat holds remain a low-time option.
Do not quote these falls to an athlete with normal pressure: larger falls went with higher baseline pressure, and no trained or normotensive group was included.
This is not treatment advice: a person with raised blood pressure or on medication agrees any exercise programme, and above all HIIT, with their physician, and exercise is added to medical care, not used instead of it.
Limitations
The authors': no subgroup analyses by sex, medication use or weekly dose when comparing modes, which may bias the mode comparison; baseline-pressure differences between trial populations may drive part of the ranking; GRADE low to moderate for most comparisons; isometric training could not be dose-modelled.
Risk of bias: 23 of 105 trials high and 53 moderate; allocation concealment and blinding were the weak domains.
Heterogeneity: most comparisons above I² 50%; for combined training vs control 95.8% (systolic) and 92.6% (diastolic).
Dose: MET-minutes come from compendium values assigned to each mode, not measured energy cost; where a trial did not report duration, the mode's average was used.
Things that do not add up: the methods define small trials as N < 15, the results as N < 20; the diastolic results say all modes reduced pressure "compared with COM" where control is meant; the text lists dynamic resistance training last for systolic pressure although its SUCRA (32.71%) is slightly above isometric (32.27%).
Not read: the supplement, so the number of isometric trials, their protocols, the league table and the meta-regression coefficients are unknown to this entry.
Funding: National Social Science Fund of China (one author); no disclosures.
Revision 2026-11-10 (decision): the lines on SYNTH-88 follow the guide as it stands — it no longer treats isometric training as the best mode, and with raised pressure it names aerobic plus strength training as the first lever and holds as the add-on when time is short, as the base's inference.
How this entry was checked and changed
Why this level
risk of bias and certainty read in the full text: 29 trials low, 53 moderate and 23 high risk of bias, GRADE mostly low to moderate (no high-certainty diastolic comparison; ranking certainty moderate for systolic, low for diastolic), and most comparisons heterogeneous (I² above 50%; 95.8% for combined training vs control, systolic), so serious enough to hold it below A