By Joshua Moss, MD

Associate Professor of Clinical Medicine, Cardiac Electrophysiology, Division of Cardiology, University of California, San Francisco

Dr. Moss reports he is a consultant for Abbott, Biosense Webster, and Boston Scientific.

SYNOPSIS: In a retrospective analysis of electronic health record data matched with remote pacemaker and implantable cardioverter-defibrillator recordings of atrial fibrillation episodes, a threshold daily arrhythmia burden portending higher stroke risk was determined over a range of CHA2DS2-VASc scores.

SOURCE: Kaplan RM, Koehler J, Ziegler PD, et al. Stroke risk as a function of atrial fibrillation duration and CHA2DS2-VASc score. Circulation 2019;140:1639-1646.

Multiple studies have demonstrated a correlation between stroke risk and the burden and duration of atrial fibrillation (AF) episodes detected by implantable pacemakers and defibrillators. Kaplan et al sought to explore further this correlation and its interaction with traditional stroke risk factors, as assessed by the CHA2DS2-VASc score, in a large, retrospective study cohort. The Optum electronic health record (EHR) database, with deidentified data from patients collected between 2007 and 2017, was used in conjunction with linked data from a Medtronic CareLink database.

A total of 28,032 patients met the assigned inclusion criteria: an implanted device with an atrial lead capable of detecting AF, at least six months of CareLink device data before an assigned index date from which to calculate AF burden, at least 12 months of EHR data before that assigned index date from which to calculate a CHA2DS2-VASc score, and at least six months of EHR data after the assigned index date from which to assess rates of ischemic stroke and systemic embolism. Average patient age was 69 years, and 63% were male. Diagnoses of transient ischemic attack (TIA) were excluded, given the more ambiguous definition in the EHR. Burden of AF was classified into three groups based on maximum daily AF duration in the six months before the index date: no AF (< 6 minutes), 6 minutes to 23.5 hours, and > 23.5 hours.

In patients not on anticoagulation, both longer AF duration and a higher CHA2DS2-VASc score were significantly associated with an increased rate of stroke and systemic embolism (SSE). The rate was slightly higher in the subset of those patients with a clinical history of AF at baseline (and also recorded a higher mean CHA2DS2-VASc score). Patients with a CHA2DS2-VASc score ≥ 2 with AF detected (but not on anticoagulation) tended to be slightly older and were more likely to be female, but clinical contraindications for anticoagulation were unavailable.

There was a clear interaction between AF duration and CHA2DS2-VASc score when assessing stroke risk in patients not on anticoagulation. In patients with a CHA2DS2-VASc score of 2, SSE risk was less than 1% per year when maximum AF daily duration was < 23.5 hours, and 1.52% per year with duration > 23.5 hours. On the other hand, in patients with a CHA2DS2-VASc score of 3-4, SSE risk was less than 1% per year only when maximum AF daily duration was < 6 minutes, and 1.28-1.77% per year when > 6 minutes. Patients with a CHA2DS2-VASc score ≥ 5 had a stroke risk > 1.5% per year, regardless of AF duration (or even whether they were on anticoagulation).

The authors concluded a combination of AF duration and CHA2DS2-VASc score can be used to make more informed decisions about initiation of anticoagulation. Using a 1% yearly stroke rate as the “actionable threshold” (the threshold at which the risk of ischemic stroke exceeds the risk of intracranial hemorrhage on oral anticoagulation), Kaplan et al hypothesized a favorable effect of anticoagulation in patients with a CHA2DS2-VASc score of at least 2 when maximum AF duration exceeded 23.5 hours/day, and in patients with a CHA2DS2-VASc score of 3-4 when maximum AF duration exceeded six minutes/day.


This large study, with its robust rhythm-monitoring data, was strong in several ways, making it useful for informing discussions with patients about the risks and benefits of therapeutic anticoagulation. The criteria used to define stroke were fairly strict, which should improve the overall specificity, but might underestimate actual SSE rates. Patients and clinicians alike often wonder whether brief episodes of AF lasting less than a day truly increase stroke risk. These data provide information that is more nuanced than yes/no and that is both clinically plausible and simple to convey in layman’s terms.

There also are some caveats that must be considered. First, regarding accuracy and applicability of the data, a study of EHR and CareLink databases comes with all the usual pitfalls of a retrospective analysis. Outcomes in patients with pacemakers and implantable cardioverter-defibrillators may not be directly generalizable to the entire population. Interestingly, the overall rate of SSE in patients not on anticoagulation (615 strokes in 21,768 patients, about 2.8%) was similar to that of patients who were on anticoagulation (175 strokes in 6,264 patients, about 2.8%). Events in patients on anticoagulation were clustered in those with a CHA2DS2-VASc score ≥ 5, while more than 15% of events in the absence of anticoagulation occurred in patients with a CHA2DS2-VASc score of 0-1 (and more than one-third recorded a score of ≤ 2).

Additionally, calculated stroke risks would have favored anticoagulation based on the hypothesized “actionable threshold” of 1% in all patients with a CHA2DS2-VASc score ≥ 5, regardless of whether they experienced AF (the event rate was 1.51% in anticoagulated patients with a high risk score and no AF). All this raises the question of whether the AF is predominantly a causative factor of SSE, or simply another marker of risk. That question will be particularly relevant when thinking about how much anticoagulation reduces stroke in high-risk patients, and deciding whether episodic rhythm-guided use of anticoagulation is ever appropriate (since events may not be temporally related to episodes of AF). Prospective trials that may help answer these questions are ongoing.

Finally, Kaplan et al stated that patients “at intermediate clinical risk may benefit from continuous AF burden monitoring to refine their individualized stroke risk.” Inevitably, there will be attempts to extrapolate these findings to commercially available devices, such as the Kardia monitor and Apple Watch. However, considering their inferior sensitivity and specificity when compared to an implanted device with an atrial lead, such extrapolation must be performed with caution.