504 CHAPTER 15 Atrial Fibrillation
whether stable rotors exist in human AF and whether they are respon-
sible for maintenance of AF, rather than a passive epiphenomenon of
disorganized fibrillatory activation. The ability to determine the true
mechanism of apparently focal activation can be limited by inadequate
spatiotemporal resolution, insufficient sampling duration, low fidelity
of interpolation, and poor differentiation of true local activation from
far-field contamination. In fact, studies showed that FIRM-identified
rotor sites did not exhibit quantitative atrial electrogram characteristics
expected from rotors (i.e., greater electrical periodicity, unique spectral
characteristics, such as higher dominant frequency and higher Shannon
entropy) and did not differ quantitatively from surrounding tissue.209
Furthermore, one report demonstrated the lack of relationship between
sites of FIRM-identified rotors and low voltage areas identified on atrial
substrate maps.213 Therefore the area targeted for ablation by the software
may not be related to the pathophysiologic mechanisms underlying the
arrhythmia. Moreover, it remains unclear whether functional rotors
can be effectively ablated.209
Another possibility is the dependence of FIRM on a basket cath-
eter that may not adequately adapt to individual variation in atrial
geometry to provide optimal atrial coverage. It is inevitable that the
basket catheter will bridge endocardial valleys and fail to fill out all
the endocardial geometry. Also, panoramic mapping with the basket
catheter provides insufficient spatial resolution due to poor tissue
contact by the basket catheter electrodes, basket distortion, and inter-
spline bunching, as well as inadequate electrode density to accurately
detect rotors near the equatorial electrodes. This results in incomplete
sampling of the atrial surface, potentially leaving mechanistically criti-
cal regions unmapped. To account for the missing electrograms (up
to 50%) and limited spatial resolution of a basket catheter, extensive
interpolation is required. Even within regions with adequate physical
sampling, the signal quality is often too poor or the atrial coverage
insufficient to perform phase analysis. The resulting activation maps
are of such low resolution that it is difficult to precisely identify the
site of any localized sources for ablation. This shortcoming would be
expected to miss sources that may be present rather than falsely show
sources. In addition, there are limitations, though minor, inherent in the
representation of a complex 3-D structure as a 2-D rectangular grid of
regularly spaced electrodes that assumes an ideal spherical fit of the basket
catheter.16,209,210,214
While significant questions remain regarding existence, methods of
detection and meaning of rotors/focal sources detected by FIRM, the
relatively high rate of AF termination to either AT/AFL or NSR and
apparent improvement in mid-term freedom from recurrent AF in
larger observational studies suggest the technique has validity.
ABLATION OF COMPLEX FRACTIONATED
ATRIAL ELECTROGRAMS
Rationale
Evaluation of the complexity and frequency of intracardiac electrograms
can help in understanding the pathophysiology of AF. Atrial electrograms
during sustained AF have three distinct patterns: single potentials, double
potentials, and complex fractionated potentials. Continuous propaga-
tion of multiple wavelets in the atria and wavelets as offspring of atrial
reentry circuits has been suggested as the mechanism by which AF can
be perpetuated without continuous focal discharge. Fractionated and
continuous electrical activity has been assumed to indicate the presence
of wave collision, slow conduction, or pivot points where the wavelets
turn around at the end of the arcs of functional blocks. Thus areas
of CFAEs during AF can potentially represent continuous reentry of
the fibrillation waves into the same area or overlap of different wave-
lets entering the same area at different times, although the mapping
resolution to discern whether such rotors even exist in human AF
is still limited.
Such complex electrical activity has a relatively short CL and het-
erogeneous temporal and spatial distribution. A relatively short CL may
indicate the presence of a driver, analogous to the frequency gradient
from the drivers or rotors to the rest of the atria observed in experi-
mental models of AF, in which the central core of these rotors can have
high-frequency electrical activity, whereas the periphery of the rotors
displays complex electrograms because of wave break and fibrillatory
conduction. Importantly, regions of CFAEs have been shown to remain
spatially and temporally stable in individual patients, particularly when
these regions were measured over several seconds, a surprising finding
considering the earlier observation that a proposed underlying mecha-
nism for AF is random reentry and that the reentrant wavelets are
expected to meander.
Several studies suggested that areas of CFAEs are critical sites for
AF perpetuation and can serve as target sites for AF ablation. It has
been proposed that once CFAEs are eliminated by ablation, AF can no
longer be sustained because the random reentry paths are altered or
eliminated so that the fibrillation wavelets can no longer reenter the
ablated areas. Whereas PV isolation targets the triggering foci, ablation
of CFAEs targets the substrate for AF. However, some connections
between both approaches probably exist. Data suggest that PVs and
their antra are the key areas where CFAEs are located; these areas need
to be ablated to achieve conversion of AF to NSR. It is thus very likely
that many patients may respond to ablation in the PV regions because
of both trigger elimination and substrate modification.
Mapping Complex Fractionated Atrial Electrograms
Mapping of CFAEs is performed during AF. In patients in NSR at the
time of study, AF is induced by rapid atrial pacing, with or without
isoproterenol infusion. During sustained AF, biatrial electroanatomic
mapping is performed. The CS or RA appendage recording is used for
electrical reference during mapping. Atrial CLs are monitored and
recorded from the reference and mapping catheters. Sites with CFAEs
can be tagged and associated with the atrial geometry created by the
electroanatomic mapping system and thereby serve as target sites for
ablation. During AF, the local activation time of the arrhythmia is of
no value in guiding activation sequence mapping.
The definition of CFAEs has not been consistent. Initially, CFAEs
were defined as: (1) fractionated atrial electrograms composed of at
least two deflections or have a perturbation of the baseline with con-
tinuous deflection of a prolonged activation complex over a 10-second
recording period; or (2) atrial electrograms with a very short CL (less
than 120 milliseconds) averaged over a 10-second recording period
(Fig. 15.49). More recently, CFAEs have been defined as having any of
the following: (1) a magnitude of less than 0.25 mV; (2) a duration
longer than 50 milliseconds with multiple (greater than 3) deflections
from the isoelectric line; or (3) continuous electrical activity without
an isoelectric line, verified visually. Some investigators, using unipolar
mapping, defined fragmented potentials as electrograms exhibiting two
or more negative deflections within 50 milliseconds. The limit of 50
milliseconds was based on the assumption that the atrial refractory
period during AF and the interval between two successive fibrillation
waves were 50 milliseconds or more.30,215
An important limitation of this approach is that the visual appear-
ance of CFAEs is variable, and they often are of very low amplitude
(less than 0.25 mV); therefore their identification by visual inspection
can be challenging and is highly subjective and investigator-dependent
(Fig. 15.50). To improve the accuracy of CFAE mapping, custom software
has been developed with algorithms that enable automated detection
and tagging of areas of CFAEs and incorporated into either the CARTO
 CHAPTER 15 Atrial Fibrillation 505

C omplex fractionated atrial electrograms

II

V1

LA appendage

LA roof

LA septum

His

C S prox

C S dist
400 msec
Fig. 15.49 Examples of Complex Fractionated Atrial Electrograms. Fractionated atrial electrograms with
a very short cycle length, compared with the rest of the atria, were recorded in the left atrial (LA) roof. In
the LA septum, fractionated electrograms with continuous prolonged activation complex were observed.
See text for discussion. CSdist, Distal coronary sinus; CSprox, proximal coronary sinus.

I
II
III
V1
His prox
His dist
C S prox
C S dist

Abl prox

Abl dist
400 msec

Fig. 15.50 Fractionated Atrial Electrograms. Recordings from near ligament of Marshall (Abl) showing
complex fragmented atrial electrograms (high-frequency, somewhat repetitive spikes [arrows]); these are
not consistently present over the duration of the recording, whereas similar activity is intermittently seen in
the His recordings (right atrial). Abldist, Distal ablation site; Ablprox, proximal ablation site; CSdist, distal coronary
sinus; CSprox, proximal coronary sinus; Hisdist, distal His bundle; Hisprox, proximal His bundle.

or NavX EnSite electroanatomic mapping systems. This software offers
valuable advantages in the detection, quantification, and regionalization
of CFAEs.
Currently, the automatic algorithms are variable and are dependent
on the recording technology (NavX and CARTO). In the CARTO system,
fractionation is generally determined by the shortest complex interval
between two consecutive deflections over a measuring period. The low-
voltage window is usually programmed at 0.05 to 0.15 mV (to exclude
both noise and high-voltage signal from the analysis). The intervals
between successive peaks falling within this window are measured.
Intervals falling within a programmable duration (usually 60 to 120
milliseconds) are identified and the number of such intervals during
the entire 2.5-second sampling window calculated; this is designated
the “interval confidence level” (ICL). The assumption is that a greater
number of short intervals between low-amplitude multi-deflection
complexes in a given time duration (2.5 seconds) reflects more frequent
506 CHAPTER 15 Atrial Fibrillation
and repetitive CFAE and, hence, higher ICL. The CARTO software can
also measure CFAE by either the “shortest complex interval” (which is
the shortest interval found in milliseconds out of all the intervals identi-
fied between consecutive CFAE complexes) or the “average complex
interval” (which is the average of all the intervals identified between
consecutive CFAE complexes for each 2.5-second electrogram). CFAE
areas are displayed on the whole-chamber map in a color-coded manner
according to the degree of fractionated signals and their CLs for easier
identification (eFig. 15.21).215–217
The algorithm embedded in the EnSite mapping system uses a dif-
ferent principle, the “CFAE mean.” The CFAE mean is defined as the
average time interval between consecutive deflections (−dV/dt) in a
local AF intracardiac bipolar electrogram recorded over a specified
length of time (5 to 8 seconds). The mean interdeflection time interval
(i.e., the mean CL) is then projected onto the LA anatomic shell as a
color-coded display. The shorter the mean CL, the more rapid and
fractionated the local electrogram. Regions with mean CLs shorter than
120 milliseconds are considered to correspond to CFAE. To optimize
algorithm accuracy, bipolar recordings are filtered at 30 to 500 Hz (to
avoid sensing noise), electrogram width is set at less than 10 to 20
milliseconds (to avoid detecting far-field signals), and electrogram
“refractory” period is set at 30 to 50 milliseconds (less than 30 milli-
seconds is regarded as nonphysiological). In addition, the baseline signal
noise level is determined, and the peak-to-peak electrogram amplitude
detection limit is set just higher than the noise level (typically, 0.03 to
0.05 mV) to minimize noise detection while allowing detection of CFAEs,
which are typically of very low amplitude (less than 0.5 mV). This
algorithm is probably most analogous to the “average complex interval”
map from the CARTO software. Advantages of the EnSite system include
an adjustable duration of recording from 1 to 8 seconds and the ability
to record electrograms from multiple poles of several catheters simul-
taneously, which can potentially occupy a significant proportion of
local AF CL.196,215–217
Importantly, assessment of fractionated electrograms during AF
requires a recording duration of at least 5 seconds at each site to obtain
a consistent fractionation and accurate analysis. A small mapping catheter
tip (4 mm), good atrial contact, and stable mapping catheter position
for several seconds while mapping at each location are important to
obtain high-quality recordings.
In general, CFAEs can be identified in most (80%) areas of the LA,
but they appear to be predominantly located in the interatrial septum,
LA roof, posterior wall, mitral annulus, and PV ostia. Less commonly,
CFAEs are located in the RA, involving the septal region, crista termi-
nalis, and CTI, as well as the CS os. Notably, patients with paroxysmal
AF appear to have more CFAE sites identified around the PV ostia,
whereas CFAEs detected in patients with persistent AF appear to be
more evenly distributed over all areas of the LA. However, these findings
have not been consistent across studies using different CFAE detection
methods.
Target of Ablation
Atrial ablation is performed at atrial sites that harbor CFAEs character-
ized by the following: (1) low-voltage (range 0.04 to 0.25 mV) signals
that have multiple potentials with continuous deflection of a prolonged
activation complex; (2) stationary CFAEs that have temporal and spatial
stability; and (3) short CL (less than 120 milliseconds) electrograms
that occur repeatedly with a relatively stable frequency with or without
multiple potentials as CFAEs. CFAE sites that are fleeting, have high
amplitude, or have a relatively long CL (greater than 150 milliseconds)
are not targeted by ablation.218
The atrial septum, followed by the regions of the PVs, is the most
common site for CFAEs. The most common localizations for termina-
tion and regularization of AF during CFAE ablation are the regions of
the PV ostia, the interatrial septum, and the LA anterior wall close to
the roof of the LAA.
After ablation of CFAEs in the LA, those in the CS and RA are
targeted. RA ablation aimed at termination of AF can potentially offer
a clinical benefit in patients with longstanding persistent AF, but it does
not seem to improve outcome in patients with persistent AF of shorter
durations.218
Ablation Technique
The ablation typically begins at sites at which CFAEs have the shortest
interval and preferably also have a high ICL. RF energy is applied using
an 8-mm-tip or, preferably, an irrigated-tip catheter. Lower power output
is applied in the CS and along the LA posterior wall. RF application is
typically continued for 30 to 60 seconds at each site, aiming to eliminate
or organize local electrograms.196
The typical response of CFAE ablation procedures for persistent AF
is a progressive increase in CL, and AF organizes into AFL or AT. It is
rare for persistent AF to convert to NSR without changing to AT or
AFL first.218
Once AF organizes into AFL or AT, residual CFAEs around the tagged
or previously ablated areas are sought and ablated. When the areas with
CFAEs are completely eliminated, the focus or reentry circuit underlying
the AT is mapped and specifically targeted by ablation. Administration
of IV ibutilide can help lengthen the tachycardia cycle length (TCL),
reduce residual fibrillatory conduction, and unmask the primary arrhyth-
mia. If the arrhythmias are not successfully terminated by ablation or
ibutilide, external cardioversion is performed.218
When CFAE ablation is combined with PV isolation, some investiga-
tors recommend that CFAE ablation be performed before PV isolation,
since the PV antra commonly harbor target CFAE. Others, however,
recommend targeting CFAEs after, rather than before, PV isolation
because that latter strategy can itself reduce the burden of CFAEs and
minimize the consequent need for extensive LA ablation.196,218
Endpoints of Ablation
The ablation endpoints when targeting CFAEs are uncertain. In most
studies, the primary endpoints were complete elimination of the areas
with CFAEs or organization and slowing of local electrograms, conver-
sion of AF to NSR (either directly or first to an AT) for patients with
persistent AF, or noninducibility of AF (with isoproterenol and atrial
pacing) for patients with paroxysmal AF. When areas with CFAEs are
completely eliminated, but arrhythmias continue as organized AFL or
AT, those arrhythmias are mapped and ablated.30
AF termination (conversion to AT or NSR) in patients with persistent
AF can potentially be a challenging endpoint to achieve and generally
requires very long procedure times. Reports demonstrated the limited
ability of CFAE ablation to terminate persistent AF. Furthermore,
although AF termination during ablation can potentially predict the
mode of recurrence (AT vs. AF), its correlation with long-term success
is still controversial. AF recurrence rates of more than 50% were observed
even in patients in whom AF was terminated during CFAE ablation.
Thus ablation of all areas displaying the electrogram of interest in the
LA and CS can be a reasonable alternative endpoint when ablation fails
to terminate AF.
Preliminary reports suggested the clinical utility of monitoring
dominant frequency in real time to guide catheter ablation of AF. A
critical decrease (11% or higher, measured in lead V1 and the CS) of
dominant frequency following CFAE ablation can potentially indicate
adequate elimination of drivers of AF and is associated with clinical
efficacy that is as high as when AF is terminated by ablation. However,
these findings require validation in prospective studies.
 CHAPTER 15 Atrial Fibrillation 506.e1

P aroxysmal A F

1.00
RAO PA

RS RS
LA A LS

RI RI
FO 7.00 LI
MV
A
P ersistent A F

1.00
RAO PA
RS LS
LA A RS

RI
RI
FO 11.00
MV LI
B
eFig. 15.21 CARTO Maps of Complex Fractionated Atrial Electrograms (CFAEs). CFAE maps with reg-
istered left atrial (LA) CT surface reconstruction (shown as wire frame), shown in right anterior oblique (RAO)
view (left-sided images) and posteroanterior (PA) view (right-sided images), in patients with paroxysmal atrial
fibrillation (AF) (A) and persistent AF (B). The CFAE maps are color-coded, with red representing the highest
interval confidence level (ICL) and purple representing the lowest ICL. Highly repetitive CFAE sites (ICL
greater than or equal to 5) are tagged with light blue dots on the CFAE maps. As shown in this example,
the highly repetitive CFAE sites are more likely located at the pulmonary vein (PV) ostia, interatrial septum,
and mitral annulus area in patients with paroxysmal AF, whereas patients with persistent AF have highly
repetitive CFAE sites predominantly identified on the LA posterior wall. FO, Fossa ovalis; LAA, left atrial
appendage; LI, left inferior PV; LS, left superior PV; MV, mitral valve; RI, right inferior PV; RS, right superior
PV. (From Scherr D, Dalal D, Cheema A, et al. Automated detection and characterization of complex fraction-
ated atrial electrograms in human left atrium during atrial fibrillation. Heart Rhythm. 2007;4:1013.)
 CHAPTER 15 Atrial Fibrillation 507

exists across studies for uncertain reasons. Whether all CFAE sites need
Outcome to be targeted for catheter ablation is still not known, and reliable
Initial single-center studies using CFAE ablation as a stand-alone strategy criteria to distinguish active from passive electrogram patterns and
for ablation of AF showed high success rates (up to 92% freedom from define optimal ablation targets are lacking.221
AF at 1-year follow-up after one or two ablation procedures). However,
these results were not universally reproducible by other investigators, PULMONARY VEIN DENERVATION
and several recent trials demonstrated that CFAE ablation alone is not
a sufficient strategy for successful treatment of AF. Rationale
On the other hand, CFAE ablation can be of potential value as an Experimental and clinical data suggest that the autonomic nervous
adjunct strategy in combination with PV isolation, particularly in patients system plays a critical role in the initiation and maintenance of AF.
with persistent AF whose response to other ablation strategies is sub- High-frequency stimulation of epicardial autonomic plexuses can induce
optimal, as well as patients undergoing redo ablation for recurrent AF. triggered activity from the PVs and potentially shorten the atrial refrac-
However, even this approach has come under considerable scrutiny tory periods, providing a substrate for the conversion of PV firing into
with the publication of randomized studies and meta-analyses showing sustained AF. Clinical studies found that ablation of the ganglionated
no additional benefit conferred by CFAE ablation in conjunction to plexuses located at the antra of the PVs (by specifically targeting the
PV isolation as compared to PV isolation alone in patients with per- ganglionated plexuses or inadvertently during standard PV ablation
sistent or longstanding AF. Furthermore, in patients undergoing PV procedures) can potentially reduce the risk of recurrence of AF.41
isolation plus CFAE ablation, there was no significant difference in
outcome between those who had complete CFAE elimination and those Localization of Ganglionated Plexuses
who did not.146,196,219,220 Autonomic inputs to the heart converge at several locations; these con-
Notably, the mode of recurrence after CFAE ablation was predomi- vergence points are typically embedded in the epicardial fat pads and
nantly organized AT or AFL, usually related to gaps within the ablation form ganglionated plexuses that contain autonomic ganglia and nerves.
regions. Commonly, extensive CFAE ablation creates islands of noncon- There are seven major ganglionated plexuses, including four located in
tiguous lesions, leading to areas of slow conduction and predisposing to the LA around the PVs and one located within the ligament of Marshall.
atrial macroreentry, thereby increasing the proarrhythmic potential.30 In the LA, the ganglionated plexuses are located around the antral
The conflicting results of different studies can be partly the result regions of the PVs and in the crux (Fig. 15.51). The superior left gan-
of variability of mapping techniques, the inconsistencies in CFAE elec- glionated plexus is located on the roof of the LA, medial to the left
trogram interpretations, differences in the type and size of the mapping superior PV, and often extends to the medial aspect of the LAA. The
electrode, and differences in the accompanying lesion set.219 Furthermore,
multiple algorithms and visual methods were used in different studies
to guide CFAE ablation, which likely resulted in targeting different
“substrate” sites, resulting in variable outcomes. In fact, in a recent
study evaluating the diagnostic accuracy of several automated algorithms
against manual determinations, CFAEs determined by automated bipolar
algorithms were highly variable and correlated poorly with established
AF substrate complexity measures and the agreement among them was
poor. Furthermore, results of those algorithms were sensitive to inter-
electrode spacing with an increase in CFAE detected with increased
interelectrode distance.216
Importantly, the benefits of AF substrate ablation have to be bal-
anced against the potential risks in individual patients, especially given
that CFAE ablation involves additional ablation lesions that can encom-
pass extensive amounts of the atrial surface area, which can potentially
compromise long-term atrial contractile function, produce arrhyth-
mogenic atrial scarring, prolong procedure and radiation times, and
increase risks of acute procedural complications.
In its current iteration, the CFAE ablation approach is limited by
uncertainties regarding the mechanistic significance, efficacy, and the
endpoint of the ablation strategy. The lack of standardization in CFAE
definition and differences in mapping technologies and electrogram
measurements, especially given the lack of comparative information
for the current technologies, adds to the challenge. Furthermore, although
multiple studies have found that ablation at sites of CFAEs could prolong
the CL or terminate AF, their true significance in the pathophysiology
of AF remains to be determined, and the sensitivity and specificity of Fig. 15.51 Left Atrial (LA) Ganglionated Plexuses (GPs). Schematic
CFAEs in identifying sites critical to perpetuation of AF are uncertain. posterior view of the LA. The five major left atrial autonomic GPs and
axons (superior left GP, inferior left GP, anterior right GP, inferior right
The mechanistic relevance of all sites of CFAEs in specific clinical con-
GP, and ligament of Marshall) are shown. Hatched areas represent the
texts and in individual patients can potentially differ. Some CFAE sites anterior LA surface, and solid areas represent the posterior LA surface.
can simply result from passive atrial activation and reflect shortening The GPs innervate the adjacent pulmonary veins and surrounding atrium.
in the AF CL, random collision of fibrillatory waves, wave disruption Interconnecting neurons connect within and between GP. (From Stavrakis
adjacent to rapidly firing foci or rotors, or nonuniform anisotropic S, Nakagawa H, Po SS, et al. The role of the autonomic ganglia in atrial
conduction. In fact, significant spatiotemporal variability of CFAEs fibrillation. JACC Clin Electrophysiol. 2015;1:1–13.)