Exercise modalities and dose for improving arterial stiffness in risk-enriched adults: a systematic review and dose-response network meta-analysis

Background Exercise is recommended for cardiovascular prevention, but the comparative effects of different exercise modalities and weekly doses on carotid-femoral pulse wave velocity (cfPWV) in heterogeneous risk-enriched adults with cardiovascular risk-related conditions or characteristics remain uncertain. This systematic review and dose-response network meta-analysis aimed to compare exercise modalities for improving cfPWV and to explore model-derived dose-response patterns and approximate favorable dose regions. Methods PubMed, Cochrane Library, Web of Science, Embase, SPORTDiscus, and CINAHL were searched from inception to 22 February 2026 for English-language parallel-group randomized controlled trials reporting cfPWV in heterogeneous risk-enriched adults with cardiovascular risk-related conditions or characteristics. Pairwise meta-analysis, frequentist network meta-analysis, and Bayesian dose-response network meta-analysis were performed. Results Thirty-seven studies involving 2,110 participants were included. In direct pairwise analysis, exercise was associated with a reduction in cfPWV compared with control (WMD = -0.65, 95% CI -1.13 to -0.18), but heterogeneity was considerable (I² = 94.4%). Aerobic training and whole-body vibration training showed significant reductions versus control in modality-specific direct comparisons, whereas resistance training, combined training, interval training, and isometric handgrip training did not reach statistical significance. In the network meta-analysis, aerobic training (WMD = -0.65, 95% CI -1.08 to -0.22) and interval training (WMD = -1.07, 95% CI -1.66 to -0.49) were associated with lower cfPWV than control. The dose-response model suggested a nonlinear association, with the lower model-derived crossing region around 200–300 MET·min/week and the lowest fitted estimate around 700–800 MET·min/week at the overall exercise level. However, these values are approximate model-derived ranges, not exact clinical thresholds, and should be interpreted cautiously because of clinical heterogeneity, sparse evidence in some dose regions and exercise nodes, and uncertainty around threshold estimates. Conclusions Exercise may reduce cfPWV in heterogeneous risk-enriched adults with cardiovascular risk-related conditions or characteristics, particularly through aerobic and interval-based training in the network synthesis. Direct pairwise evidence supported exercise overall, aerobic training, and whole-body vibration training, but the certainty of evidence was generally low or very low and heterogeneity was considerable. The dose-response findings suggest a potential moderate-dose region with lower model-predicted cfPWV estimates at the overall exercise level, but these estimates should be considered model-derived and hypothesis-generating rather than definitive clinical prescription thresholds. These findings should be interpreted as average effects across diverse populations and exercise prescriptions rather than as uniform expected effects. Trial registration PROSPERO CRD420261372601.