The meta-analysis by Lafuente-Lafuente et al1 of treatments for maintaining sinus rhythm addresses a key aspect of the management of atrial fibrillation (AF), but the authors failed to deal with the enormous heterogeneity in the designs of the studies.
Patients with chronic atrial fibrillation differ substantially from those with the paroxysmal condition in their ability to maintain sinus rhythm, yet the populations studied varied regarding the type of AF and even in the definitions of “paroxysmal” and “chronic” used to classify patients. They also differ on the methodological issue of whether patients need to be in sinus rhythm at the time of randomization. Thus, pooling the data assumes that patients who were in sinus rhythm for several months after a paroxysm of AF2 are comparable with those cardioverted only a couple of hours prior to randomization.3 Moreover, some studies do not count early recurrences2,4 and some allow patients to be cardioverted again, while others count every recurrence no matter how soon it occurs.
Caro JJ, Xenakis JG, Huybrechts KF. Ignoring Study Design Heterogeneity in Meta-analyses. Arch Intern Med. 2006;166(18):2046. doi:10.1001/archinte.166.18.2046-a