Atrial fibrillatory cycle length (AFCL) is generally accepted as a surrogate marker for local refractoriness. In this study, a computer model and clinical data on human subjects undergoing catheter ablation for paroxysmal and persistent AF were used to determine the clinical potential of AFCL.
Simulations were performed in a biophysical computer model of AF, induced from eight simultaneously active focal sources. Atrial fibrillatory cycle length persistence and termination were assessed in response to successively switching off the involvement of the eight sources. Electrophysiological data were obtained from 178 subjects undergoing catheter ablation of AF. Atrial fibrillatory cycle length, measured in the atria appendages using automated monitoring software, was studied to determine its clinical correlation, the complexity of the ablation procedure, and the AF termination success rate. Computer simulations showed an inverse relationship between the number of sources participating in AF maintenance and AFCL. Clinical data demonstrated a strong relationship between duration, degree of ablation, and AFCL, with shorter AFCL associated with more extensive ablation to terminate AF. Atrial fibrillatory cycle length was prolonged exponentially at each stage, with a critical cycle length of ∼200 ms for AF conversion.
Atrial fibrillatory cycle length is inversely associated with the number of sources participating in AF maintenance observed in the computer model. In addition, AFCL is an important predictor of baseline duration of the arrhythmia, type of AF, and ease of catheter ablation therapy to terminate AF.
Atrial fibrillation (AF) is the arrhythmia most commonly encountered in clinical practice. The clinical spectrum ranges from paroxysmal to persistent and chronic. The chronicity of AF is associated with progressive electrical remodelling, 1–3 shortening of atrial refractoriness, and reduction in the likelihood of restoration and maintenance of sinus rhythm. 4 Different electrophysiological mechanisms may be involved in perpetuating AF. Classically, these include the presence of multiple wandering wavelets, and macro-re-entrant loops. More recently, activities/driving sources emanating from the thoracic veins and/or atrial tissue have been demonstrated. Biophysical computer modelling of human atria may provide insight into the underlying mechanisms and elements involved in AF perpetuation that cannot be studied using conventional mapping. 5
The atrial fibrillatory cycle length (AFCL), expressing the fibrillatory rate, is generally accepted as a surrogate marker for local refractoriness. It has been used to evaluate responses to anti-arrhythmic drugs and to electrical cardioversion. The value of using AFCL to monitor the progress of atrial ablation therapy has recently been demonstrated; gradual prolongation of the AFCL has consistently been observed to precede the termination of arrhythmia during the ablation of paroxysmal or permanent AF. 6 , 7
In this study, AFCL was measured in a computer model of AF and compared with data obtained from human subjects undergoing catheter ablation for paroxysmal and persistent AF with the aim of determining its clinical potential. A computer model of AF based on multiple sources was used. The multiple source hypothesis postulates that a small number of sources (be it foci or rotors), presumably located in the left atrium (LA), maintain fibrillatory activity. 8 Here, such sources were simulated by periodically injecting intracellular current into a group of cells. A series of experiments were performed to investigate the number of sources involved in AF perpetuation/maintenance and termination and their contribution.
Multiple sources model for atrial fibrillation
Biophysical computer models of AF require the specification of the substrate properties of the tissue (geometry, cell, and conduction properties) and the electrical stimulation protocol. Such computer models of AF have been described earlier. 5 , 9–11 The computer model used in this study was based on a segmented human magnetic resonance imaging data set, used to generate a three-dimensional description of the atrial geometry. 12 The wall thickness ranged from 1.2 to 1.5 mm, resulting in a mesh comprising ∼800 000 nodes with a spatial resolution of 0.3 mm. Conduction properties were taken to be uniform (70 cm/s), except in the fast conduction bundles that were included in the model (Bachmann's bundle, crista terminalis, and pectinate muscles, 100–120 cm/s). In order to create a substrate for AF, patchy heterogeneities in action potential duration (APD) were introduced by modifying the local membrane properties, resulting in an APD 90 distribution of 195 ± 15 ms (range 150–230 ms). 12 , 13
Atrial fibrillation could be initiated by pacing in regions exhibiting a large APD gradient and cycle lengths close to the effective refractory period. However, no such episode of simulated AF was sustained longer than 20–30 s. In order to simulate long-lasting AF episodes, multiple sources were introduced within the atrium. Eight locations were selected in the LA, mainly in the pulmonary vein region. These locations, shown in Figure 1 , represent typical targets for source ablation in patients. The electrical activity of these focal sources was simulated by periodically injecting intracellular current into the region. The cycle lengths of sources 1–8 were 160, 180, 155, 163, 175, 168, 172, and 200 ms, respectively. Atrial fibrillation was initiated by switching on the eight sources simultaneously. Complex activity patterns developed after ∼1 s in the LA and after 8 s in the right atrium (RA).
Simulation study protocol
Atrial fibrillation was first triggered by the simultaneous introduction of the eight focal sources in the LA. Ten seconds later, the impact of the individual sources was extinguished sequentially by creating a non-propagating region around them. The resulting impact on AFC and continuation or termination of AF was studied.
Nine steps were defined. At step 0, the eight sources were active for 10 s. At step 1, the first source was extinguished and the simulation was continued for 5 s, and similarly, at step 2, the second source was extinguished and the simulation was continued for 5 s. This process was continued as far as step 8, at which source 8 was extinguished and the simulation run until self-termination of AF occurred.
Local AFCL values over the 10 s periods were derived from the action potentials simulated at 5042 points uniformly distributed over the entire right and LA surface. For each step, maps were constructed ( Figure 2 ), displaying the result of the statistical testing of the change in the local AFCL values relative to those of the preceding step. Colour code was the following: green if no significant variation ( P > 0.05) in AFCL was observed, red for a significant reduction in AFCL ( P < 0.05), and blue for a significant increase in AFCL ( P < 0.05).
A total of 68 patients with paroxysmal AF and 178 patients with persistent or permanent AF undergoing catheter ablation were enrolled in the study. Atrial fibrillation had been persistent for 25 ± 42 months (median 12, range 1 month to 36 years). All anti-arrhythmic medications with the exception of amiodarone ( n = 37) were ceased prior to the ablation procedure. Oral anticoagulation was administered for at least 1 month prior to the procedure, aiming at a target international normalized ratio between 2 and 3. Transoesophageal echocardiography was performed in all patients within 5 days of the procedure to exclude the presence of atrial thrombus. The study protocol was approved by the institutional Clinical Research and Ethics Committee. Written informed consent was obtained from all patients; the possible use of an investigational catheter for mapping was explained to them.
Surface electrocardiogram and bipolar endocardial electrograms were continuously monitored and stored on a computer-based digital amplifier/recorder system (Labsystem Pro, Bard Electrophysiology). Intracardiac electrograms were filtered from 30 to 500 Hz.
The following catheters were introduced via the right common femoral vein for electrophysiological study: (i) a steerable quadripolar or decapolar catheter (5 mm electrode spacing, Xtrem, ELA Medical) was positioned within the coronary sinus (CS); (ii) a 20-pole mapping catheter with either circular configuration for mapping venous perimeters (Lasso, Biosense-Webster) or arranged in five radiating limbs reaching a diameter of 3.5 cm (PentaRay, Biosense-Webster) introduced following transeptal access, and (iii) an irrigated-tip ablation catheter with a distal 3.5 mm tip and three 1 mm electrodes separated by 2-5-2 mm inter-electrode spacings (ThermoCool, Biosense-Webster). The multi-limb catheter, with its high density electrode configuration, was used in five patients to map atrial regions in an attempt to identify the earliest site of centrifugal activation.
Following trans -septal puncture, a single bolus of 50 IU/kg of heparin was administered and repeated only for procedures lasting ≥4 h if the activated clotting time (ACT) was <200 s. Following pulmonary vein isolation, the circular catheter was removed from the sheath and left in the RA appendage and the irrigated-tip catheter advanced via the same sheath for LA and CS ablation.
Staged ablation approach
The ablation of the LA required up to four steps: isolation of pulmonary veins (PVI), ablation of the CS region, ablation of atrial tissue, and linear LA ablation. PVI ablation was performed in the standard fashion guided by a circumferential mapping catheter as previously described. The roofline, a linear ablation line connecting the two superior PVs, 14 was performed at its utmost cranial portion using parallel and/or perpendicular catheter-tissue configurations. Ablation of the CS started along its endocardial aspect (inferior LA bordering the mitral annulus) and from within the vessel to complete the lesion, if required. Left atrium ablation was performed at all sites displaying any of the following electrogram features potentially representing arrhythmogenic tissue: continuous electrical activity, 7 complex fractionated potentials, 15 sites with a gradient of activation (significant electrogram offset between the distal and proximal recording bipoles on the map electrode) possibly indicating a local rotating wave, or regions with a cycle length shorter than the mean LA appendage AFCL (LAACL) or demonstrating centrifugal propagation to the surrounding tissue. 16 Ablation at all of these atrial sites was performed for 20–60 s to achieve local prolongation of cycle length. In patients whose AF persisted after the above steps, linear ablation at the mitral isthmus was performed with the endpoint of linear block as previously described. 17
When organization of LA activity was achieved by cumulative stepwise ablation resulting in consistent atrial activation sequences, activation mapping was performed in an attempt to identify driving sources demonstrating centrifugal propagation. This step involved sequential mapping with conventional catheters or the use of the multi-limb catheter. We have previously reported that such drivers may display local electrograms varying from discrete to complex electrical activity at their site of origin. 7
Radiofrequency (RF) energy was delivered at each site or area with a power of 20–30 W (inside venous structures) and 30–40 W (atrial tissue) using irrigation rates of 5–60 mL/min (0.9% saline via a Cool Flow pump, Biosense-Webster) to achieve the desired power delivery. Less power was delivered when the catheter tip was perpendicular to the ablation site in order to minimize the risk of popping. Temperature was limited to 50°C. The duration of RF delivery was annotated at each area.
Monitoring of atrial fibrillatory cycle length: impact of ablation
The effect of ablation was monitored as previously described by evaluation of the AFCL and the termination of AF. The effect of local ablation was assessed by the impact on mean fibrillatory cycle length, as well as by whether or not AF terminated. Mean AFCLs were determined immediately before and after each ablation step, by averaging 30 consecutive cycles using automated monitoring software (Labsystem Pro, Bard Electrophysiology) and visual verification. The electrograms within the LAA were usually distinct and of high amplitude, thereby facilitating unambiguous automatic annotation. The process of assessing cycle length required <60 s and was performed in review mode, allowing the intervention to remain uninterrupted. Atrial fibrillatory cycle length was determined within the right and LAAs. At each time point, the automated annotation was manually verified and corrected if needed, using online calipers at a paper speed of 100 mm/s. An individual site was considered to have a significant impact if its ablation resulted either in the termination of AF or in a prolongation of the AFCL (evaluated in the LAA except if specified otherwise) by 5 ms or more, when compared with the longest AFCL observed during the previous steps. Termination of AF was defined as either a transition directly to sinus rhythm or a conversion to atrial tachycardia (AT). When AF was converted to a regular arrhythmia, activation and entrainment mapping were performed to differentiate between a focal and a macro-re-entrant mechanism.
Continuous variables are reported as mean ± SD or median and range. Comparison between groups was performed with the Student's t -test or when data were not normally distributed (Shapiro–Wilk test), the Wilcoxon rank-sum or signed-rank test. Categorical variables are reported as numbers and percentages and were compared using the Fisher's exact test. Statistical significance was established at P < 0.05.
The effect of reducing the number of sources involved, as reflected in the AFCL, was as follows ( Figure 2 ): Step 0: 182 ± 42 ms; Step 1: 184 ± 46 ms; Step 2: 180 ± 39 ms; Step 3: 188 ± 44 ms; Step 4: 187 ± 40 ms; Step 5: 199 ± 44 ms, Step 6: 206 ± 58 ms; Step 7: 400 ± 3 ms (conversion to AT: cycle length of 200 ms with a 2:1 block); Step 8: sinus rhythm. Significant changes in the AFCL between steps were observed only after extinguishing the fifth focal source (199 ± 44 vs. 182 ± 42 ms, P < 0.05) indicating a delayed impact of ablation on AFCL. This can be explained from Figure 3 , which shows that despite the presence of eight firing sources only three or four sources were active at a given time, presumably because these individual sources were sufficient to maintain LA refractoriness impeding manifestation of other sources. Ablation of the initial sources allowed the appearance of previously unseen sources so that significant prolongation of AFCL was not observed during the initial elimination steps. In Figure 2 , the significant increase in local LAACL is highlighted in blue. Conversion of AF into AT was observed after extinguishing the seventh source, with a preceding AFCL of 206 ± 58 ms. Termination of AT to sinus rhythm followed the removal of the final focal source.
Atrial fibrillatory cycle length in paroxysmal vs. persistent atrial fibrillation
The values of AFCL in paroxysmal vs. persistent AF differed significantly (170 ± 14 vs. 156 ± 23 ms, P = 0.002). Patients receiving prior amiodarone therapy had higher AFCL in both patient groups (171 ± 27 vs. 148 ± 17 ms, P < 0.01). The presence of structural heart disease was associated with a longer AFCL (159 + 23 vs. 149 + 22 ms, P = 0.003). Among patients with persistent AF, there was an inverse correlation between the duration of AF and AFCL ( Figure 4 ). The fibrillatory rate estimated from the surface electrograms during AF showed a good correlation with the invasively determined AFCL in a subset of 77 patients from this cohort. A significant correlation was found between the surface AFCL and both of the LAACL and the RAACL ( R = 0.901, P < 0.001 and R = 0.820, P < 0.001, respectively). The surface AFCL had a strong linear correlation with the duration of the procedural time ( R = −0.640, P < 0.001). Additionally, the surface AFCL was significantly longer in patients in whom AF terminated by ablation alone as compared to patients without termination AF (AFCL 152.2 ± 17.2 vs. 135.7 ± 11.5 ms, P < 0.05).
Baseline atrial fibrillatory cycle length and termination of persistent atrial fibrillation during ablation
Radiofrequency delivery resulted in termination of AF in 149 of 178 patients (85%), directly into sinus rhythm in 23/149 (15%) or into intermediate AT in 126/149 (85%). Atrial fibrillation could not be terminated in 29 patients (15%) after completion of the study protocol. Such patients demonstrated shorter AFCL in the LAA at baseline (141 ± 20 vs. 156 ± 22 ms, P < 0.001) and less prolongation of AFCL in the LA following ablation (27 ± 12 vs. 35 ± 9 ms; P = 0.06). In all patients, termination of AF was preceded by an average increase in the AFCL (from 156 ± 22 to 196 ± 32 ms, P = 0.0001). Atrial fibrillation terminated during LA ablation in the majority of patients 133 (84%) patients, whereas additional ablation in the RA terminated AF in an additional 16 patients. In the former, left atrial ablation increased AFCL in a similar way for both RA and LA, whereas in the latter, left atrial ablation had a reduced impact on AFCL prolongation in the RA, resulting in a left–right fibrillatory cycle gradient ( Figure 5 ), suggesting that the driver for AF was in the RA.
Multivariate analysis incorporating age, gender, structural heart disease, amiodarone use showed baseline AFCL to be the strongest independent predictor of AF termination in patients with chronic AF (CAF) duration of <5 years ( Figure 6 ). However, in patients with CAF duration >5 years, the rate of AF termination with catheter ablation was modest (44%) with AFCL having no predictive value and the duration of CAF appeared to be the strongest predictor of successful AF termination on multivariate analysis.
Baseline atrial fibrillatory cycle length and extent of ablation to terminate atrial fibrillation
The total RF ablation duration delivered to atrial or venous tissue was 87 ± 27 min. The total procedure and fluoroscopic times were 255 ± 66 and 86 ± 28 min, respectively. The ease of terminating AF by ablation, as indicated by the duration of the procedure duration or total RF delivery, varied inversely with baseline AFCL. The extent of ablation inferred from the duration of RF delivery was higher for shorter AFCL ( Figure 7 ).
This study presents new information on the determinants of AFCL, using computer simulation and results from catheter ablation. In the computer simulations, AFCL was found to be inversely related to the number of sources/elements participating in the AF process (higher number of elements involved correlating with shorter AFCL). Clinical studies corroborated this finding, as demonstrated by a strong relationship between duration, degree of ablation, and AFCL, with shorter AFCL associated with a more extensive ablation procedure required to terminate AF. Prolongation of AFCL rises exponentially with each stage of ablation, saturating at AF conversion at (a ceiling for) the critical cycle length of ∼200 ms.
Computer simulations of atrial fibrillatory cycle length
Most of the described computer simulations involve multiple re-entrant wavelet or heterogeneous tissue models. In a computer simulation study of multiple wavelets, Rotter et al . 11 demonstrated that termination of AF occurred more frequently with a larger excluded surface area and with the addition of linear lesions. This result agrees with previously described clinical studies 18 showing cumulative effects of ablation on termination of AF and further confirm the utility of computer simulation model to aid in our understanding of AF. Likewise, the computer simulation model in this study, used to determine AFCL and its relationship with a number of firing sources and the impact of source elimination, provide important insights. Progressive prolongation of AFCL was observed prior to termination when the majority of the sources was eliminated and was accompanied by an organized form of activity within the atrium. At any given time, only part of the eight computed drivers were apparently active, the others being quiescent. This could be explained by refractoriness of the atrium allowing only the manifestation of three or four sources at a particular time instant. Consequently, when initial sources are extinguished or abolished, their contribution is replaced by other firing sources, and the impact on global AFCL will appear later when the majority of the sources are extinguished.
Clinical importance of atrial fibrillatory cycle length
Atrial fibrillatory cycle length can be clinically evaluated through the basic frequency of the f-wave as observed on the surface ECG or, invasively, by the recording of atrial electrograms on atrial tissue during electrophysiological procedures. Measurement of the f-wave frequency on the surface ECG requires both sufficient amplitude of atrial activity and a slow ventricular rate to avoid superimposition of QRS waves on atrial activity. QRS subtraction and amplification of atrial electrograms or their analysis in the frequency domain can be performed to identify more clearly the fibrillatory rate. 19 , 20 Higher f-wave frequencies are observed in permanent AF or in the absence of drug therapy. However, the f-wave frequency on the surface ECG correlated more with RA than LA activity, whereas the latter is demonstrated in humans to be the dominant chamber for AF perpetuation. This is supported by our findings, which showed a higher predictive value of LAACL for successful AF termination compared with RAACL.
Intracardiac measurement of AFCL requires sharp electrograms to provide unambiguous annotation of cycle length, and these are best obtained from the atrial appendages. In other regions, fractionation of electrograms or continuous activity impedes identification of the exact electrogram timing. In our patients, we observed a strong inverse relationship between AFCL and the duration of sustained AF; AFCL was shorter in persistent than paroxysmal AF confirming prior studies and was shorter for the longer lasting AF in the persistent AF group. This phenomenon is linked to functional changes in atrial tissue properties (‘remodelling’) demonstrated experimentally in many species. This is characterized consistently by a shortening of the atrial action potentials and the refractoriness seen after weeks, up to months, of rapid electrical activation without influence of the RA or LA location of stimulation.
A strong relationship was observed between the baseline AFCL (prior to any ablation) and the outcome of ablation. Atrial fibrillatory cycle length was significantly longer in patients in whom AF terminates during ablation compared with those patients in whom AF persists. The likelihood of termination was indeed proportional to AFCL, with the best cut-off being 140 ms.
In addition, the gradual prolongation of AFCL consistently observed during ablation had a greater magnitude in patients with AF termination than those without. Once a critical prolongation of the AFCL has been achieved, usually in the region of 180–200 ms, the atrium could no longer sustain the fibrillatory process and AF terminated either by conversion directly to sinus rhythm or, in the majority of cases, to AT. This critical AFCL associated with AF termination was a remarkable clinical finding, which explains that by starting at 130 ms it would require a longer road (ablation extent) to reach 200 ms than by starting, for example, at 170 ms. Thus, it is not surprising that the facility to terminate AF by ablation, as indicated by the procedure duration or total RF delivery, varied inversely with the baseline AFCL. Importantly, AFCL is less predictive of procedural outcome in AF lasting >5 years, during which period extensive atrial remodelling and fibrosis may have modified the number and characteristics of the participating elements.
A tentative hypothesis
Atrial fibrillatory cycle length represents the local activation rate derived from local atrial electrograms and is mainly considered as an indicator of the local refractory period. The fact that the refractory period is not the only determinant of AFCL is confirmed by the prolongation of AFCL during ablation at remote sites. On the basis of the present study, we hypothesize that AFCL may be the result of the summation of activities/elements contributing simultaneously to AF at any given time in the site measured (e.g. activities converging to the appendage), with the lower limit of AFCL being the local refractory period. A short AFCL (<140 ms) would indicate rapid activity of some elements or multiplicity of these elements and short refractory period; in contrast, a long AFCL may express either smaller number/slower rate of the agents driving AF or longer refractory periods of atrial tissue (partial remodelling). A progressive suppression of elements by ablation, simulated or real, may initially have little impact on AFCL because of contributions from other elements at the measured site. When most of the contributory elements have been eliminated, there is the proportionately higher rise in AFCL, culminating in AF termination.
The term ‘elements’ is taken in a broad sense without differentiating the different types of re-entry or sources involved in maintaining AF. The number of elements, the rate and also the duration of their intrinsic activity can impact on AFCL, which is the final common denominator. This hypothesis integrates both the ensemble of AF mechanisms at play in a given patient and the extent of the remodelling process leading to shortening of refractoriness. The challenge for biophysicists will be to frame the properties of multiple elements in terms of variables that can be tested in computer models in order to gain insights into the clinical understanding and relevance of AFCL.
In practice, when AFCL is short, the implication is that multiple influences are at work, and therefore, a more extensive intervention is needed to treat the arrhythmia and presumably a therapy is less likely to interrupt AF. This is demonstrated for all therapies presently available now (anti-arrhythmic drugs or effectiveness of cardioversion or catheter ablation). Conversely, when AFCL is long, fewer elements may be contributing to fibrillation and therefore will lead to a more limited procedure for restoring normal rhythm.
Atrial fibrillatory cycle length is a readily accessible parameter, by invasive and non-invasive methods, that may have several clinical applications. First, it could be a useful index to enable prediction of AF duration and type, its correlation with structural remodelling in atria and response to pre-treatment anti-arrhythmic medications. Secondly, AFCL appears to be an important clinical tool in selecting candidates for catheter ablation, to tailor and guide individualized therapy and optimize the results of ablation. A very short AFCL (<140 ms), predicting difficult and extensive ablation, may indicate the need for a pre-treatment with amiodarone to prolong AFCL and thus facilitate ablation.
In the computer model, only one theory of AF was tested in ideal electrophysiological conditions. No difference between the LA and RA was included, in spite of clear dominance of LA in human AF. In addition, factors such as intra-atrial conduction or autonomic system were not assessed.
In the clinical study, AFCL variations were tested at a single site in the atrium, which is probably more influenced by the closer activities and may not record all the activities at play in the entire chamber. The results of this study are based in part on the effect of ablation in different LA regions, resulting in AFCL prolongation and subsequently AF conversion. These results do not indicate the absence of influence on the fibrillatory process at sites where no prolongation of AFCL was observed.
This study was made possible by grants from the Theo-Rossi-Di-Montelera Foundation, Medtronic Europe, the Swiss Governmental Commission of Innovative Technologies (CTI), and the Swiss National Science Foundation (SNSF).
The authors thank Ryan Lahm and Drs. Josée Morisette and Arthur Stillman who kindly furnished the atrial geometry surface model and Prof. Adriaan van Oosterom for his editorial assistance.
Conflict of interest : Dr. Nathalie Virag is a full time employee of Medtronic.