Four RCTs were included in the review (6,064 participants, 5,967 included in analysis; range 102 to 4,628). Two of the RCTs reported in a common publication and generated a single effect estimate. Overall the risk of bias was low (data not reported).
When a fixed-effect analysis was used, dronedarone was associated with a significantly lower risk of stroke or transient ischaemic attack than placebo, without heterogeneity (RD -0.0094, 95% CI -0.0178 to -0.0011; four RCTs, Χ² p=0.3). In this model one study, the ATHENA trial (4,628 participants), had 79.5% of the statistical weight.
When a random effects analysis was used, there was no significant difference between the groups (RD -0.0064, 95% CI -0.0144 to 0.0016; four RCTs,) and the ATHENA trial had 45.1% of the statistical weight and the combined EURIDIS/ADONIS trials (1,237 participants) had 50.5% of the statistical weight.