Antidepressants and Atrial Fibrillation: The Importance of Resourceful Statistical Approaches to Address Confounding by Indication

Atrial fibrillation (AF) is the most common cardiac arrhythmia. Antidepressant drugs are known to influence cardiac conduction, and there are theoretical grounds to suggest that serotonin reuptake inhibition by antidepressants may predispose to AF. A large register-based cohort study found that antidepressant use was associated with an increased risk of AF during the first month following antidepressant initiation. Whereas this may seem alarming, additional analyses found that the risk progressively diminished in succeeding months. Further analysis showed that antidepressant use was associated with an even higher risk of AF in the month before antidepressant initiation; this finding implies that patients who initiate antidepressant drugs are already at increased risk of AF and that the significant association between antidepressant drug use and AF might be because of confounding by indication. In other words, the indication for antidepressant use rather than antidepressant use, itself, may be the risk factor for AF, and this risk is probably state-dependent because it diminishes with the passage of time, during which period the indication for antidepressant use presumably resolves. In another study, which used a nested case-control design, current and recent users of antidepressant drugs were contrasted not with antidepressant nonusers but with past users of antidepressants. The study found that neither current nor recent users were at increased risk of AF. This again indicates that antidepressants are unlikely to directly predispose to AF. In both studies, the possibility of confounding by indication was addressed by resourceful statistical approaches to the research question. Such resourcefulness in analysis is necessary in all observational studies of the association between an exposure and an outcome. In the absence of efforts to address confounding by indication, an identified association between an exposure and an outcome should not be presumed to reflect a cause-effect relationship.