Exercise-based management of patients with coronary heart disease: an integrated network meta-analysis and machine learning exploration

How exercise interventions should be structured to improve peak oxygen uptake (VO₂peak) in coronary heart disease (CHD) remains uncertain. This study aimed to assess the relative effects of different exercise modalities on VO₂peak, identify study-level moderators of effect, and characterise dose-response relationships using meta-regression models. We systematically reviewed and synthesised 129 randomised controlled trials involving 7,760 participants with CHD. A network meta-analysis (NMA) was used to estimate the relative effects of various exercise modalities. To investigate heterogeneity, a meta-analytic random forest (MARF), interpreted with Shapley Additive Explanations (SHAP), was employed to identify influential study-level moderators. Meta-regression models were used to examine dose-response relationships. Most exercise modalities were associated with improvements in VO₂peak relative to the control group. Interval aerobic exercise training (mean difference [MD] = 3.43; 95% credible interval [CrI]: 2.87, 3.99) and combined aerobic and resistance exercise training (MD = 2.49; 95% CrI: 1.81, 3.17) showed relatively larger estimated effects. Dose-response analyses indicated that greater weekly exercise dose was generally associated with larger improvements in VO₂peak, whereas per-session exercise dose showed an inverted-U-shaped association. MARF and SHAP analyses highlighted intervention length as an influential study-level moderator. Exercise-based cardiac rehabilitation improves VO₂peak in adults with CHD. However, substantial heterogeneity and low-certainty evidence limit confidence in modality-specific rankings. Longer intervention duration and greater weekly exercise dose were associated with larger improvements in VO₂peak, suggesting that exercise dose and programme duration should be considered alongside patient characteristics, comorbidity burden, and safety when designing cardiac rehabilitation programmes. These exploratory and aggregate-level findings require prospective validation before guiding individualised exercise prescriptions.