When multiple treatments are available, network meta-analysis can evaluate data to rank the relative effectiveness. We applied this approach to first-line treatments for paroxysmal atrial fibrillation (medical therapy, radiofrequency ablation or cryoballoon ablation). Individual trials were analysed based on the restricted mean survival time (RMST). Randomised controlled trials (RCT) assessing first-line treatments for paroxysmal atrial fibrillation were referenced from PubMed and the websites of regulatory agencies. The primary end-point was atrial fibrillation recurrence-free survival at 12 months. The treatments assessed for their relative effectiveness were medical therapy, radiofrequency ablation and cryoballoon ablation. Individual trials were examined based on RMST. A Bayesian network meta-analysis was conducted to comparatively evaluate these treatments. Five trials were included in the analysis: two compared radiofrequency with medical treatment and three cryoballoon ablation with medical treatment. The indirect comparison of radiofrequency ablation vs cryoballoon ablation was assessed in the absence of RCTs. Differences in RMST (with 95% credible intervals) were estimated for all binary comparisons (direct or indirect). Radiofrequency and cryoballoon ablation showed significantly increased effectiveness compared with medical treatment. In the indirect comparison, radiofrequency showed a non-significant advantage over cryoballoon ablation. The ranking of effectiveness was as follows: (1) radiofrequency; (2) cryoballoon ablation; (3) medical treatment. In conclusion, we found that radiofrequency was the most effective treatment for paroxysmal atrial fibrillation according to a Bayesian probabilistic model.