We included 20 RCTs involving a total of 2039 participants with AF. All trials were conducted between 2006 and 2024, with a follow-up period ranging from eight weeks to five years. We assessed the certainty of evidence as moderate to very low. Five trials assessed comprehensive ExCR programmes, which included educational or psychological interventions, or both; the remaining 15 trials compared exercise-only cardiac rehabilitation with controls. The overall risk of bias in the included studies was mixed. Details on random sequence generation, allocation concealment, and use of intention-to-treat analysis were typically poorly reported.

Evidence from nine trials (n = 1173) suggested little to no difference in mortality between ExCR and non-exercise controls (risk ratio (RR) 1.06, 95% confidence interval (CI) 0.76 to 1.49; I² = 0%; 101 deaths; low-certainty evidence). Based on evidence from 10 trials (n = 825), ExCR may have little to no effect on SAEs (RR 1.30, 95% CI 0.63 to 2.67; I² = 0%; 28 events; low-certainty evidence). Evidence from four trials (n = 378) showed that ExCR likely reduced AF recurrence (measured via Holter monitoring) compared to controls (RR 0.70, 95% CI 0.56 to 0.88; I² = 2%; moderate-certainty evidence). ExCR may reduce AF symptom severity (mean difference (MD) −1.59, 95% CI −2.98 to −0.20; I² = 61%; n = 600; low-certainty evidence); likely reduces AF symptom burden (MD −1.61, 95% CI −2.76 to −0.45; I² = 0%; n = 317; moderate-certainty evidence); may reduce AF episode frequency (MD −1.29, 95% CI −2.50 to −0.07; I² = 75%; n = 368; low-certainty evidence); and likely reduces AF episode duration (MD −0.58, 95% CI −1.14 to −0.03; I² = 0%; n = 317; moderate-certainty evidence), measured via the AF Severity Scale (AFSS) questionnaire. Moderate-certainty evidence from six trials (n = 504) showed that ExCR likely improved the mental component summary measure in health-related quality of life (HRQoL) of the 36-item Short Form Health Survey (SF-36) (MD 2.66, 95% CI 1.22 to 4.11; I² = 2%), but the effect of ExCR on the physical component summary measure was very uncertain (MD 1.75, 95% CI −0.31 to 3.81; I² = 52%; very low-certainty evidence). ExCR also may improve individual components of HRQoL (general health, vitality, emotional role functioning, and mental health) and exercise capacity (peak oxygen uptake (VO _{2 峰} ) and 6-minute walk test) following ExCR. The effects of ExCR on serious adverse events and exercise capacity were consistent across different models of ExCR delivery: centre compared to home-based, exercise dose, exercise only compared to comprehensive programmes, and aerobic training alone compared to aerobic plus resistance programmes. Using univariate meta-regression, there was evidence of significant association between location of trial and length of longest follow-up on exercise capacity.