Effectiveness of various exercise interventions on peak VO2 and quality of life in patients with heart failure with preserved ejection fraction (HFpEF): a network meta-analysis

Fig. 1. PRISMA flowchart of the study selection process. Flowchart illustrating the systematic literature search and study selection process following PRISMA guidelines. Records were identified from PubMed, Embase, Web of Science, and the Cochrane Library. After duplicate removal, title/abstract screening, and full-text assessment, 14 RCTs were included in the network meta-analysis

Fig. 2. Risk of bias assessment. Panel A Summary bar chart showing the proportion of studies rated as low risk (green), unclear risk (yellow), or high risk (red) for each domain of the Cochrane Risk of Bias Tool (RoB 1.0). Panel B Individual risk of bias judgments for each included study. Rows represent individual studies; columns represent the seven bias domains: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other bias. Plus signs (+) indicate low risk, question marks (?) indicate unclear risk, and minus signs (−) indicate high risk

Fig. 3. Network structure diagrams. Panel A Evidence network for peak VO2. Panel B Evidence network for quality of life (MLWHF scores). Each node (blue circle) represents one intervention, and node size is proportional to the total number of participants assigned to that intervention across included trials. Connecting lines indicate direct head-to-head comparisons, and line thickness is proportional to the number of studies contributing to each comparison. The number of studies and total sample size are displayed beside each node

Fig. 4. Forest plots for peak VO2 outcomes. Panel A All pairwise comparisons among interventions from the network meta-analysis for peak VO2 (mL/kg/min). Each row represents a comparison between two interventions. Panel B Mean difference in peak VO2 for each intervention compared to no exercise (reference). In both panels, diamonds represent point estimates (mean differences) and horizontal lines represent 95% confidence intervals. Positive values indicate greater improvement in peak VO2 favoring the first-named intervention. Panels A and B are complementary: Panel A provides the complete comparison matrix, while Panel B isolates the effect of each intervention relative to a single reference

Fig. 5. Forest plots for quality of life outcomes (MLWHF Scores). Panel A All pairwise comparisons among interventions from the network meta-analysis for MLWHF scores. Each row represents a comparison between two interventions. Panel B Mean difference in MLWHF scores for each intervention compared to usual care (reference). In both panels, diamonds represent point estimates (mean differences) and horizontal lines represent 95% confidence intervals. Negative values indicate greater reduction in MLWHF scores (i.e., better quality of life) favoring the first-named intervention. Panels A and B are complementary: Panel A provides the complete comparison matrix, while Panel B isolates the effect relative to a single reference

Fig. 6. SUCRA cumulative ranking plots. Panel A Surface under the cumulative ranking curve (SUCRA) plots for peak VO2. Panel B SUCRA plots for quality of life (MLWHF scores). Each subplot displays the cumulative ranking curve for one intervention. SUCRA values range from 0% (certainly the worst) to 100% (certainly the best); higher values indicate a greater probability of the intervention being among the most effective. SUCRA values are displayed within each subplot

Fig. 7. Rankograms (Ranking Probability Plots). Panel A Ranking probabilities for improving peak VO2. Panel B Ranking probabilities for improving quality of life (MLWHF scores). The x-axis indicates rank position (Rank 1 = most effective; higher numbers = less effective), and the y-axis indicates the probability of achieving that rank. Each colored line represents one intervention (HIIT, MICT, IMT, usual care, and no-exercise control). Ranking probabilities sum to 1.0 for each intervention. Curves concentrated near Rank 1 indicate a higher likelihood of being the best treatment

Fig. 8. League tables (Pairwise Comparison Matrices). Panel A All pairwise comparisons for peak VO2 (mL/kg/min). Panel B All pairwise comparisons for MLWHF scores. Each cell shows the mean difference (95% CI) comparing the row intervention to the column intervention. In Panel A, positive values favor the row intervention (greater improvement in peak VO2). In Panel B, negative values favor the row intervention (greater reduction in MLWHF scores, indicating better quality of life)

Fig. 9. Comparison-adjusted funnel plots for publication bias assessment. Panel A Funnel plot for peak VO2. Panel B Funnel plot for MLWHF scores. The x-axis represents the effect size centered at the comparison-specific pooled effect; the y-axis represents the standard error (inverted scale). Each point represents one study-specific comparison. Dashed lines indicate pseudo 95% confidence limits. Symmetrical distribution around the vertical line suggests low likelihood of publication bias