Effects of exercises based on ACSM recommendations on patients with heart failure with preserved ejection fraction: a systematic review and meta-analysis of randomized controlled trials

Figure 1. Flow diagram of literature selection.

Figure 2. (A) Risk of bias graph based on the Cochrane risk-of-bias tool. Seven bias domains were assessed: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other bias. Bars represent the percentage of included trials judged as low risk, unclear risk, or high risk of bias for each domain. (B) Risk of bias summary graph based on the Cochrane risk-of-bias tool. Horizontally, it lists 7 criteria for bias assessment: random sequence generation, group concealment, blinding of subjects and researchers, blinding of outcome assessment, incomplete outcome data, selective reporting, and other biases; vertically, it shows 19 included studies (divided into 20 groups), presenting intuitively the individual risk assessment results of each study in each bias dimension.

Figure 3. Forest plot of SMD for 6-minute walk test (6-MWT) stratified by intervention compliance (random-effects model). Intervention significantly improved 6-MWT performance only in the high-compliance subgroup; subgroup difference was statistically significant, with high interstudy heterogeneity (I2=90%).

Figure 4. Forest plot of SMD for peak oxygen consumption (peak VO2) stratified by intervention compliance (random-effects model). Intervention significantly elevated peak VO2 in both subgroups; efficacy was stronger in high-compliance group, with significant subgroup difference (P=0.01). Overall high interstudy heterogeneity (I2=81%).

Figure 5. Forest plot of SMD for the total score of the Minnesota Living with Heart Failure Questionnaire (MLHFQ total score) stratified by intervention compliance (random-effects model). Intervention significantly reduced MLHFQ scores only in the high-compliance subgroup; no significant subgroup difference was observed, with overall moderate-to-high interstudy heterogeneity (I2=77%).

Figure 6. Forest plot of SMD for left ventricular ejection fraction (LVEF) stratified by intervention compliance (fixed-effects model). Intervention showed significant positive effect only in the high-compliance subgroup; no significant inter-subgroup difference was found, and no interstudy heterogeneity existed (I2=0%).

Figure 7. Forest plot of SMD for the ratio of mitral early diastolic inflow velocity to mitral annular early diastolic velocity(E/e’ radio) stratified by intervention compliance (fixed-effects model). Neither subgroup exhibited significant intergroup difference; overall pooled analysis indicated a mild significant benefit of intervention, with no statistical discrepancy between subgroups and low interstudy heterogeneity (I2=30%).

Figure 8. Forest plot of SMD for early diastolic transmitral flow velocity / Atrial systolic transmitral flow velocity ratio(E/A radio) stratified by intervention compliance (fixed-effects model). No significant intergroup difference was observed in either subgroup or overall pooled analysis; no statistical discrepancy existed between subgroups, and no interstudy heterogeneity existed (I2=0%).

Figure 9. Forest plot of SMD for ventilation / Carbon Dioxide production slope(VE/VCO2 slope) stratified by intervention compliance (random-effects model). No significant intergroup difference was found in two subgroups and overall pooled analysis; no difference existed between subgroups, with moderate-to-high interstudy heterogeneity (I2=72%).

Figure 10. Forest plot of SMD for deceleration time(DT) stratified by intervention compliance (fixed-effects model). No significant intergroup difference was detected in both subgroups and overall pooled analysis; subgroup difference was non-significant, and overall interstudy heterogeneity was zero (I2=0%).

Figure 11. Funnel plots assessing publication bias across all outcomes. The plots exhibited approximate symmetry, suggesting no substantial publication bias. (A) 6-MWT; (B) peak VO2; (C) total MLHFQ score; (D) LVEF; (E) E/e′ ratio; (F) E/A ratio; (G) VE/VCO2 slope; (H) deceleration time. SMD, standardized mean difference; SE, standard error.