Key Question 1: Medical Treatment Versus Radiofrequency Ablation (RFA)

There is a moderate level of evidence to show that patients who received RFA as a second-line therapy (i.e., in patients who did not respond to medical therapy) had a higher chance of maintaining sinus rhythm compared to those treated with medical therapy alone (relative risk (RR) 3.46, 95% CI 1.97–6.09) at 12 months postprocedure. This finding is in general agreement with a previously published metaanalysis.¹⁴² We did not find a statistically significant difference in the risk of cerebrovascular events in patients who were treated with RFA compared to those treated with medical therapy. However, clinically meaningful differences could not be excluded because the event rates were small and studies were not powered to detect such small differences.

There were insufficient data to draw meaningful conclusions concerning RFA use as a first-line therapy for rhythm control (i.e., in patients who have never been treated with antiarrhythmic drugs (AADs)).

Key Question 2: Patient- and Intervention-Level Characteristics Associated with Rhythm Control

There is low level of evidence to show that atrial fibrillation (AF) type, namely nonparoxysmal AF, is predictive of a higher rate of AF recurrence. Although metaanalyses of univariable analyses support an association (RR about 1.6 suggesting more recurrence with nonparoxysmal AF), the studies were clinically and statistically heterogeneous, and more importantly, only six of 17 multivariable analyses bear this out, with hazard ratios ranging from 1.1 to 22 (favoring paroxysmal AF).

There is a moderate level of evidence to show that among patients with approximately normal ejection fraction (EF) or left atrial diameter (LAD), these parameters are not independent predictors of AF recurrence; however, there is insufficient evidence to estimate the predictive value of abnormal EF or LAD on recurrence rates. There is a high level of evidence to show that sex, AF duration, and the presence of structural heart disease are not associated with AF recurrence. Among patients between approximately 40 and 70 years of age, there is a high level of evidence to show that age is not associated with AF recurrence; however, the evidence is insufficient to estimate the predictive value of young or very old age. There is insufficient evidence for other potential predictors of AF recurrence.

Key Question 3: Approaches and Technical Issues Concerning RFA

Key Question 4: Adverse Events Associated with RFA

There is a low level of evidence suggesting major adverse events associated with RFA are relatively uncommon. The level of evidence was rated low because the studies reviewed employed nonuniform definitions and assessments of adverse events, with sample sizes generally less than 100, and incomplete reporting. While there is no doubt that certain adverse events are uniquely associated with the use of RFA (e.g., atrioesophageal fistula), the limitations cited precluded accurate estimates of those adverse event rates. Asymptomatic PV stenosis, cardiac tamponade, and cerebrovascular events were reported at rates of 4% or less in the majority of the studies. Symptomatic PV stenosis was reported at rates of 1% or less. Four studies reported rates of atrioesophageal fistula ranging from 0.1% to 0.9%. A total of five deaths were reported in all the studies reviewed (one patient died from a pulmonary infection, one died from anaphylaxis after the procedure, and three died from atrioesophageal fistulas (three publications from the same group of investigators each reported one death from atrioesophageal fistula). However, it is difficult to compare the rates of adverse events across studies as the descriptions of the various adverse events were not always comparable. For example, even though the presence of PV stenosis was generally evaluated at around 3 months post-RFA, severe and moderate PV stenoses were defined differently across studies. Some clearly reported stroke as periprocedural, and some reported stroke without stating a time of occurrence. Also, it was not always made clear whether the lack of information on a particular adverse event meant zero events (i.e., the researchers systematically ascertained for it and found none) or it was simply not spontaneously reported. In addition, the sample sizes in most RCTs and comparative studies were generally small, precluding reliable risk estimates of the adverse events. Furthermore, many of the studies had a mean followup of no more than 12 months, any long term events like late AF recurrence or mortality or delayed adverse effects from radiation exposure could not be properly assessed from this group of studies.