OBJECTIVES: Atrial fibrillation (AF) is a potentially life-threatening arrhythmia. After restoration of a normal sinus rhythm, AF recurrence may be prevented with anti-arrhythmic drugs (AADs). The objective of this review was to evaluate the efficacy of a novel AAD (dronedarone) and existing AADs with respect to mortality outcomes. METHODS: A systematic literature review was conducted of randomised clinical trials (RCTs) investigating the use of AADs in patients with AF. All‑cause mortality and CV mortality data were extracted. Direct and indirect comparisons adopted an intention-to-treat basis using Peto odds ratios (OR) with a fixed effect or Bucher model, respectively.  RESULTS: The majority of published data considered dronedarone, amiodarone, sotalol, flecainide or propafenone (32 publications). Peto ORs for all-cause mortality at 12 months with 95% confidence intervals (CI) for direct and indirect comparisons via non-active control were calculated. In comparisons with non-active control, the risks of all-cause mortality were as follows: a significant increase with sotalol (OR 2.72 [1.16, 6.38]); an increased trend with amiodarone and flecainide, and a decreased trend, with a narrow CI, with dronedarone (OR 0.85 [0.66, 1.09]). Indirect comparison demonstrated that dronedarone was associated with lower odds of all-cause mortality; this reached statistical significance versus sotalol (OR 3.2 [1.32, 7.78]). Indirect analysis of amiodarone versus dronedarone (OR 2.38 [0.80, 7.07]) is consistent with the head-to-head DIONYSOS study (calculated OR 2.32 [0.52, 10.32]). In both indirect and direct analyses, CIs for propafenone were too wide to interpret the results. Only one publication reported CV mortality data and this considered dronedarone. Therefore, meta-analyses were not possible. CONCLUSIONS: Clinical trial CV mortality data are limited. All-cause mortality does not appear to be significantly reduced via the use of AADs. Despite this, the novel AAD, dronedarone, was associated with lower all-cause mortality risk than those observed with existing AADs or non-active control.