This study was performed as a single-center retrospective cohort study. The study protocol adhered to the principles of the Declaration of Helsinki and was approved by the Institutional Review Board at Yonsei University Health System. All patients provided written informed consent for inclusion in the Yonsei AF Ablation Cohort (ClinicalTrials.gov identifier: NCT02138695). Among 3,640 patients who underwent a de novo AFCA from March 2009 to December 2020, we included 2,118 patients who underwent post-procedural isoproterenol provocation tests. Among them, 227 patients underwent a repeat AFCA (Figure 1). The study exclusion criteria were as follows: (1) permanent AF refractory to electrical cardioversion; (2) AF with valvular disease grade > 2; (3) prior cardiac surgery with a concomitant AF surgery. Before the ablation procedure, the anatomies of the left atrium (LA) and PVs were visually defined on three-dimensional (3D) computed tomography (CT) (64 Channel, Light Speed Volume CT; Brilliance 63; Philips, Best, The Netherlands). We confirmed the absence of any LA thrombi by transesophageal echocardiography, intracardiac echocardiography, or CT. All antiarrhythmic drugs (AADs) were discontinued for at least five half-lives, and amiodarone was stopped at least 4 weeks before the procedure.

All included patients underwent transthoracic echocardiography (Sonos 5500, Philips Medical System, Andover, MA, USA or Vivid 7, GE Vingmed Ultrasound, Horten, Norway) before undergoing AFCA and at 1 year after the procedure. We acquired the cardiac chamber size, left ventricular ejection fraction, trans-mitral Doppler flow velocity, the ratio of the early diastolic peak mitral inflow velocity, and early diastolic mitral annular velocity (E/Em), following the American Society of Echocardiography guidelines.

We recorded the intracardiac electrograms using the Prucka CardioLab™ Electrophysiology system (General Electric Medical Systems, Inc., Milwaukee, WI, USA) and generated 3D electroanatomical maps (NavX, Abbott, Inc., Chicago, IL, USA; CARTO system, Biosense Webster, Diamond Bar, CA, USA) using a circumferential PV-mapping catheter (AFocus, Abbott, Inc., Chicago, IL, USA; Lasso, Biosense-Webster Inc., Diamond Bar, CA, USA) through a long sheath. The 3D geometries of the LA and PVs were generated using the 3D mapping system and then merged with 3D spiral CT images. Blinded to the patient information, a technician analyzed the color-coded CT-merged voltage maps. Then, we performed transseptal punctures. The number of transseptal punctures (single vs. double) was at the discretion of the operator. Afterward, we obtained multi-view pulmonary venograms for the perfect matching of 3D-map, CT, and fluoroscopy in all patients except for significant renal disease. Immediately after the transseptal puncture, systemic anticoagulation was started with an intravenous bolus of heparin 200 IU/kg followed by intermittent boluses to maintain an activated clotting time of 350–400 s.

The details of the AFCA technique and strategy were described previously (12). An open-irrigated tip catheter (Celsius, Johnson & Johnson Inc., Diamond Bar, CA, USA; NaviStar ThermoCool, Biosense Webster Inc., Diamond Bar, CA, USA; ThermoCool SF, Biosense Webster Inc.; ThermoCool SmartTouch, Biosense Webster Inc.; Coolflex, Abbott Inc., Minnetonka, MN, USA; 30–35 W; 47°C; FlexAbility, Abbott Inc.; ThermoCool SmartTouch, Biosense Webster Inc., and TactiCath, Abbott Inc.) was used for the AFCA. Because we included patients over a relatively long period for this study, the radiofrequency power for the AFCA varied between 25 and 60 W. The endpoint of ablation at each site was as the average impedance drop >10% of baseline or an >80% decrease in the local electrogram voltage amplitude. We generated a CPVI with a bidirectional block in all patients. Most patients (91.6%) underwent the creation of cavotricuspid isthmus block during the de novo procedure unless there was atrioventricular conduction disease. Empirical linear ablation, including a roofline, posterior inferior line (posterior box lesion), and anterior line, left lateral isthmus ablation, right atrial ablation, or complex fractionated electrogram ablation were performed at the discretion of the operator.

After completing the protocol-based ablation, AF or atrial tachycardia (AT) was induced by 10-s high-current burst pacing (10 mA, pulse width 5 ms, Bloom Associates, Denver, CO, USA) from the high right atrial (RA) electrodes. This commenced at a pacing cycle length of 250 ms and was gradually reduced to 120 ms as previously described in the procedure (14). We infused isoproterenol (5–20 μg/min depending on ß-blocker use with a target heart rate of 120 bpm) for at least 3 min before induction and maintained this for 3 min after the induction of AF or AT. If sustained AF or AT was induced, internal cardioversion was performed by utilizing biphasic shock (2–20 J) with R wave synchronization (Lifepak12, Physiocontrol Ltd., Redmond, WA, USA). We conducted all the procedures under conscious sedation but induced deep sedation immediately before electrical cardioversion. We ended the procedure when there was no immediate recurrence of AF within the 10 min after the isoproterenol infusion with or without cardioversion. If further AF triggers were observed under the isoproterenol effect, we determined the potential location of the extra-PV triggers based on the contact bipolar electrograms and conducted a quick and detailed 3D-activation mapping with a multielectrode catheter. Based on the 3D mapping of the non-PV foci, we ablated those foci with 35–50 W for 10 s in each lesion until elimination. After the first round ablation, we performed the provocation procedure if it was a highly reproducible extra-PV trigger. However, we did not conduct the second time isoproterenol provocation in general. We defined the extra-PV foci as AF triggering points by isoproterenol provocation after the bidirectional block of CPVI.

We discharged the patients without taking any AADs except for those who had recurrent extra-PV triggers after the AFCA procedure, symptomatic frequent atrial pre-mature beats, non-sustained atrial tachycardia, or early recurrence of AF on telemetry during admission. Patients visited the outpatient clinic regularly at 1, 3, 6, and 12 months and then every 6 months or whenever symptoms occurred. All patients underwent ECG during every visit and 24-h Holter recordings at 3 and 6 months and every 6 months for 2 years, annually for 2–5 years, and then biannually after 5 years. Holter monitoring or event monitor recordings were obtained when patients reported palpitations suggestive of an arrhythmia recurrence. We defined an AF recurrence as any episode of AF or AT of at least 30 s in duration. Any ECG documentation of an AF recurrence within a 3-month blanking period was diagnosed as an early recurrence, and an AF recurrence occurring more than 3 months after the procedure was diagnosed as a clinical recurrence. We analyzed the clinical recurrences depending on the existence of extra-PV triggers and their locations and risk factors.

A GE Marquette MARS 8000 Holter analyzer (General Electric Medical Systems, Chicago, IL, USA) was used to analyze heart rate variability (HRV) based on the 24-h Holter monitor recordings. Pre-mature ventricular beats, pre-mature atrial beats, and electrical artifacts were excluded from the analysis. The mean heart rate and the following time-domain HRV parameters were analyzed as follows: mean RR interval (mean NN interval), SD of NN intervals, SD of the 5 min mean of NN intervals, and root mean square of differences between successive NN intervals (rMSSD). The following parameters were calculated as follows: very-low-frequency components (<0.04 Hz), low-frequency components (LF; 0.04–0.15 Hz), high-frequency components (HF; 0.15–0.4 Hz), and LF:HF ratio. The HF and rMSSD were indicators of parasympathetic nervous activity.

Continuous variables were summarized as the M ± SD and compared by independent two-sample t-test analysis. Categorical variables were summarized as the number (percentage of the group total) and compared by either the chi-square test or Fisher's exact-test. Multivariable logistic regression was applied to identify predictors associated with the existence of extra-PV triggers. Kaplan–Meier analysis with the log-rank-test was used to calculate AF recurrence-free survival over time and to compare recurrence rates according to the existence of extra-PV triggers. A multivariable Cox regression analysis was performed to identify the predictors associated with a clinical recurrence of AF. A two-sided P < 0.05 was considered statistically significant. The statistical analyses were performed using R version 4.0.2 software (The R Foundation, www.R-project.org, Vienna, Austria).

Table 1 summarizes the baseline clinical characteristics of the patients with extra-PV triggers during the de novo and repeats AFCA procedures. Regarding de novo procedures, the patients with an extra-PV trigger tended to be women and older and to have a longer AF duration, lesser history of vascular disease, higher LA volume indices on the echocardiogram, lower mean LA voltage, and lower pericardial fat volume. Details of procedural complications are presented in Supplementary Tables 1, 2. In the multivariable logistic regression, an older age [odds ratio (OR) 1.02 per 1-year increase, 95% CI 1–1.03, p = 0.021], no history of vascular disease (OR 0.53, 95% CI 0.3–0.89, p = 0.023), and a higher LA volume index (OR 1.01, 95% CI 1–1.02, p = 0.015) were independently associated with the existence of extra-PV triggers during the de novo procedure (Table 2). Among the 247 patients who demonstrated an extra-PV trigger in the de novo procedure, 19.8% (49/247) underwent a repeat ablation.

The median interval between the de novo and repeat procedures was 2 (interquartile range 1.1–4.1) years. The patients with extra-PV triggers in repeat procedures were more likely to be women (p = 0.014) and have diabetes (p = 0.035, Table 1). Being a woman (OR 2.72, 95% CI 1.39–5.38, p = 0.004), having diabetes (OR 2.42, 95% CI 1.07–5.47, p = 0.033; Model 1), and having higher rMSSD (OR 1.02, 95 CI 1–1.04, p = 0.025; Model 2) were independently associated with the existence of extra-PV foci during the repeat procedure (Table 2).

The proportion of patients who had extra-PV triggers at the end of the procedure was 11.7% (247/2,118) during the de novo procedure and 28.6% (65/227) during the repeat procedure (p < 0.001). The number of extra-PV trigger foci was significantly higher in the repeat ablation (1.49 ± 0.73 per patient) than in de novo procedure (1.22 ± 0.46 per patient, p < 0.001; Supplementary Table 3). The location of the extra-PV triggers in the de novo and repeat ablation procedures are presented in Figure 2. The septum (19.9%), coronary sinus (14.7%), and superior vena cava (11.2%) were the most common sites for extra-PV triggers regardless of whether a de novo or repeat procedure (Supplementary Table 3). Multifocal extra-PV triggers were documented in 21.5% of patients and were more common in the repeat procedures (18.2% in de novo vs. 33.8% in repeat ablations, p = 0.006; Supplementary Table 3). Extra-PV triggers were unmappable in 6.5 and 9.2% of the patients in the de novo and repeat ablation procedures, respectively.

Figure 2B displays the locations of the extra-PV foci in the repeat procedures in patients who showed in both de novo and repeat ablations. Table 3 summarizes the characteristics of the extra-PV triggers provoked in repeat procedures. Among the 65 patients revealed to have extra-PV triggers in redo procedures, 6.2% (4/65) had an origin at the same site as the extra-PV foci in the de novo procedure, while 84.6% (38/55) had new origins of extra-PV triggers. Among the 46 patients who were newly found to have mappable extra-PV triggers in the redo ablation, 15 (32.6%) patients had extra-PV trigger sites matched with the previous empirical extra-PV ablation sites, whereas 31 (67.4%) had extra-PV trigger sites which were not touched at de novo procedures.

Table 4 summarizes the rhythm outcomes after de novo and repeat procedures according to the existence of extra-PV foci. AADs were maintained at discharge and 3 months after the repeat procedures more frequently in patients with extra-PV foci than in those without. During the mean follow-up of 50.2 ± 37.7 months, clinical recurrences of AF were significantly higher in the patients with extra-PV triggers (44.9%) than in those without (31.6%) after the de novo AFCA (Log-rank p < 0.001, Figure 3A). During a mean follow-up of 38.8 ± 29.5 months after the repeat ablation procedure, the rhythm outcome was consistently worse in patients with extra-PV triggers (57.5%) than in those without (36.7%) (Log-rank p < 0.001, Figure 3B). The comparison between the patients with a single extra-PV trigger and multiple extra-PV triggers is presented in Figure 3C (Log-rank p = 0.073). In the Cox regression analyses (Table 5), women [hazard ratio (HR) 1.33 95% CI 1.09–1.61, p = 0.004], persistent AF (HR 1.49, 95% CI 1.25–1.75, p < 0.001), increasing LA diameter (HR 1.04, 95% CI 1.02–1.06, p < 0.001), and the presence of extra-PV triggers (HR 1.91, 95% CI 1.54–2.38, p < 0.001) were independently associated with a clinical recurrence of AF after the de novo AFCA. The best cut-off value of LA diameter predicting recurrence was ≥44 mm (Supplementary Figure 1). The presence of extra-PV triggers (HR 2.68, 95% CI 1.63–4.42, p < 0.001) and persistent AF (HR 1.82, 95% CI 1.1–3.03, p = 0.019) were associated with an AF recurrence after the repeat procedure (Table 5).

In this single-center, retrospective cohort study, we analyzed the isoproterenol-induced extra-PV triggers during de novo and repeat AF ablation procedures. Extra-PV triggers were more commonly found in the repeat ablation than in the de novo procedure. Older age and LA remodeling were independently associated with the presence of extra-PV triggers in the de novo ablation and being a woman, having diabetes, and higher parasympathetic nerve activity were related to the presence of extra-PV triggers in the repeat procedure. One-third of the extra-PV triggers newly found in the repeat ablation matched the empirical extra-PV ablation sites in the de novo procedure, and the existence of extra-PV triggers was independently associated with a higher AF recurrence after both the de novo and repeat ablation procedures.

Continuous long-term AF recurrence after AFCA is a current unmet need. Representative mechanisms of AF recurrence after ablation are PV reconnections, extra-PV triggers, and autonomic neural effects. We expect that the long-lasting PVI issue will be overcome with the development of an upgraded and effective catheter technology (15). However, the extra-PV trigger issue has not only an elusive mechanism but also has limited treatment methods. In this study, we found a significant relationship between an extra-PV trigger and atrial remodelings, such as age and LA volume. The study of Keita Watanabe et al. (16) showed that being a woman, a lower body mass index (BMI <23.8 kg/m²), absence of hypertension, and ventricular diastolic dysfunction were independent predictors of extra-PV foci in patients with paroxysmal AF. In this study, the existence of extra-PV triggers in repeat ablations was associated with women, diabetes, and cardiac parasympathetic activity as well as the previous empirical extra-PV ablation sites (32.6%). This suggested that the autonomic nervous activity and previous ablation lesions play some role in extra-PV triggers and the recurrence mechanism. In particular, myofibroblasts, ion current changes, and gap junctional remodeling accompanying matrix remodeling or local fibrosis processes lead to electrophysiological changes such as in the membrane potential, conduction velocity, and refractoriness (17, 18). This atrial substrate remodeling contributes to AF initiation and maintenance mechanisms by the perpetuation of triggers or micro re-entry. Pericardial fat volume was smaller in the patients with extra-PV triggers than those without in this study. Pericardial fat is associated with arrhythmogenicity and AF recurrence in previous studies (19, 20). However, the existence of extra-PV triggers seemed to be more closely related to being a woman rather than to pericardial fat volume or body-mass index in this study. According to the recent study regarding sex differences in the mapping of repeat ablation procedures, extra-PV triggers were more significantly frequent in women than in men. Still, pericardial fat volume was substantially smaller in women (21).

The prevalence of extra PV triggers in the de novo AFCA is variable and ranged from 3.2 to 62%, depending on the provocation protocols and punctuality of the mapping procedures (22–24). Unlike PVs, which were anatomically distinct structures, the mapping and ablation method for extra-PV triggers was elusive. A randomized-controlled trial failed to show an improvement in the rhythm outcome by additional empirical ablation for complex fractionated electrograms or rotors that were mapped during sustained AF (25). However, the study of Lee et al. reported that an extra-PV trigger ablation mapped after an isoproterenol provocation significantly lowered the AF recurrence through a randomized clinical trial (26). The work of Kim et al. also reported that the extra-PV ablation lowered the recurrence rate, but the rhythm outcome of the patients who had an extra-PV trigger was significantly worse than that of the patients without, even after an extra-PV trigger ablation (14).

There were several limitations to isoproterenol-provoked extra-PV trigger mapping and ablation. First, immediate trigger mapping was difficult, and the accuracy was decreased when the 3D map was shaken, following the electrical cardioversion due to patient movement. Second, the 3D map defined the exit site of triggers, but the actual foci might exist in the epicardial layer or deep inside the septum. Third, the isoproterenol provocation protocol has not been verified or standardized. We raised the target heart rate to 120 bpm and then induced AF considering the β-blocker effect, but it has not been proven that this dose is appropriate. In this study, there were fewer extra-PV triggers in patients with vascular disease, which might reflect the potential bias of a more careful isoproterenol dosing considering the coronary risk.

The presence of an extra-PV trigger is an important factor that must be overcome in determining long-term prognosis after AFCA. However, it was not easy to eliminate the extra-PV triggers in 11% of the de novo and 29% of the repeat procedures or 23% of multiple and 7% of unmappable foci utilizing a contact electrode catheter. Entire chamber mapping such as with the ECGi panoramic map could be a breakthrough for extra-PV mapping but has a limitation of localizing the foci on the septum, which was the most common extra-PV foci site in the current and previous studies (27). Since the AF driver map using computational modeling was an entire chamber map, it can be used as a guide for extra-PV trigger ablation (28). AADs with an appropriate dose are an alternative option to suppress extra-PV triggers by controlling the ion currents. If the presence of an extra-PV trigger could be predicted before the AFCA procedure, it would be useful in selecting an appropriate ablation catheter (balloon or RF), determining the need for an isoproterenol provocation test, and evaluating the prognosis after the procedure.

This study had several limitations. First, it was a single-center, retrospective observational cohort study that might have involved selection bias. Further multi-center, prospective studies should be conducted. Second, there was no uniform strategy for the extra PV trigger ablation, and we performed an empirical extra PV ablation based on the discretion of the operators. Finally, we included patients with an isoproterenol provocation test and excluded those who did not undergo a provocation test. This might have resulted in selection bias.