Reversible or Provoked Atrial Fibrillation?

The Devil in the Details

In a recent article published in JACC: Clinical Electrophysiology, Quon et al. (1) examined patients with “secondary” atrial fibrillation (AF), defined as new-onset AF occurring in the context of acute coronary syndromes, acute pulmonary disease, or sepsis (1). The authors did not find any association between oral anticoagulation (OAC) and ischemic stroke, but a higher risk of bleeding with OAC in acute pulmonary disease (1). Several limitations in using retrospective hospital discharge data were addressed in the accompanying editorial (2). In this letter, we highlight 2 additional problems related to this heterogeneous population.

First, although AF may develop in the context of medical illness (secondary AF), the mechanism and prognostic implications of AF can be further dichotomized into 2 distinct, yet different, populations: reversible AF, which develops solely due to acute illness, with no abnormal substrate and no future risk; and provoked AF, unmasked by acute illness in patients with abnormal substrate and ongoing risk for AF recurrence. In the article by Quon et al. (1), this distinction is most obvious in the pulmonary disease group, whereby 32% of the population had AF in the context of pulmonary infection (acute influenza or pneumonia). Influenza infection alone is associated with an 18% increased risk of developing AF, and among these patients, the benefit of long-term OAC is unclear after return to sinus rhythm (3,4). A similar analogy could be made for patients with hyperthyroidism and AF, where anticoagulation is beneficial in the hyperthyroid state but not necessarily long term (5). By contrast, provoked AF patients may have paroxysmal AF that is unmasked during acute illness. Chronic obstructive pulmonary disease patients with provoked AF may have the underlying cardiac substrate for AF and ongoing risk for AF recurrence, and thus long-term OAC may significantly reduce ischemic stroke in these patients (6). By combining reversible and provoked AF, the authors created a heterogeneous group of patients, and it is of no surprise that the benefit of OAC was unclear.

The second problem is the significant bias among patients receiving OAC in this study. Despite almost all patients meeting OAC indications based on current guidelines (all patients ≥65 years of age, 90% CHADS₂ score ≥1) (7–9), Quon et al. (1) reported a variable anticoagulation rate between 27% and 40%. This systematic and gross underutilization of anticoagulation implies unmeasured factors likely influenced the treating physicians. In one circumstance, the treating physician may have considered AF reversible and, thus, long-term anticoagulation not justified. Conversely, others might have felt that competing risks (i.e., bleeding among acute coronary syndromes patients) were not in favor of benefit (i.e., stroke reduction). In this latter case, there would be an expected regression to the mean (high bleeding risk patients and low stroke risk patients left untreated), thus appearing to show no beneficial effect of anticoagulation as well.

Finally, we do not have the absolute event rates in a similar population of patients without AF (or “denominator”) to assess the true benefit of anticoagulation. A propensity-matched analysis could delineate 2 populations with similar comorbidities, in which the impact of provoked AF and anticoagulation on ischemic stroke could be evaluated.

• 2018 American College of Cardiology Foundation