1 Introduction

For the last two decades, pulmonary vein isolation (PVI) has been the cornerstone of atrial fibrillation (AF) ablation techniques [1], evolving from an initial focal ablation inside the PVs [2] into a more safe and effective wide area approach [3]. Superior freedom from atrial arrhythmias when compared to antiarrhythmic drugs (AAD) [4, 5] alone has established PVI’s important role in AF therapy [6]. However, long-term recurrences endure as the major shortcoming of this method, ranging from 30 to 70% depending on patient’s comorbidities, type of AF, and repeat procedures [7, 8]. To overcome this limitation, research efforts have focused on technological advances and improved patient selection, for which knowledge about the predictors of recurrence is of paramount importance. Very long-term recurrence rates and whether or not to maintain AAD therapy after ablation are two of the questions that remain largely unanswered due to the scarcity of data and significant methodological differences between reports [9, 10]. Answers to these questions would be important to guide patient management and inform patient-physician discussions on what to expect after AF ablation.

The aims of this study were (1) to assess the rates and predictors of very long-term AF recurrence after a first catheter ablation procedure and (2) evaluate the impact of maintaining AAD therapy after PVI.

2 Methods

2.1 Patient population and study design

All consecutive patients with drug-refractory AF who underwent catheter ablation were included in a two-center (Hospital de Santa Cruz, Carnaxide, Portugal, and Hospital da Luz, Lisbon, Portugal) observational registry between September 2007 and December 2008. To account for the learning curve effect, the initial 30 or 50 ablations with a new catheter or navigation system, respectively, were discarded.

From a total of 361 procedures, we excluded patients with previous PVI (n = 74), those without left atrial (LA) volume assessment by computerized tomography (CT) or electroanatomical mapping (n = 23), and those who were lost to follow-up during the blanking period (n = 11), yielding a final study population of 253 individuals. LA volume has been frequently reported as one of the strongest predictors of relapse [11,12,13]; therefore, missing this information was regarded as an exclusion criterion.

Most patients (n = 180, 71%) underwent a 64-slice cardiac CT scan less than 48 h before the catheter ablation procedure for systematic evaluation of LA volume and pulmonary vein anatomy. Further integration with electroanatomical mapping was performed at the time of PVI [12]. LA volume was calculated by tracing the LA borders on CT images, excluding the pulmonary veins and the left atrial appendage. In the final study population, all patients underwent LA volume estimation using CARTO® (Biosense Webster Inc., Diamond Bar, CA, USA) electroanatomical mapping at the time of ablation. LA volume assessed by cardiac CT correlated well with 2D transthoracic echocardiography (r = 0.71, p < 0.001) and electroanatomical mapping (r = 0.79, p < 0.001) methods (Supplementary Material Fig. I). The latter was therefore used as the reference value whenever CT data was unavailable.

AF was categorized as paroxysmal if self-terminated in < 7 days, persistent whenever episodes lasted ≥ 7 days or required cardioversion, or long-standing persistent if maintained for more than 12 months [14]. Smoking was defined as any cigarette consumption during the previous 6 months. Left ventricular (LV) systolic dysfunction was defined by an ejection fraction < 50%.

All patients underwent additional transesophageal echocardiography for thrombi exclusion immediately before the catheter ablation procedure.

2.2 Pulmonary vein isolation protocol

PVI was guided either by CARTO® or NavX® (St. Jude Medical® Inc., St. Paul, MN, USA) electroanatomical mapping systems. The right femoral vein was commonly used as the preferred vascular access. A decapolar catheter was advanced through the coronary sinus; a variable circular mapping catheter placed in the pulmonary veins and a 3.5-mm irrigated-tip catheter was used for ablation [Navistar® Thermocool® (Biosense Webster Inc., Diamond Bar, CA, USA)]. LA access was established by transseptal puncture. Radiofrequency ablation was performed more than 4 mm away from the PV ostia, with continuous lesions enclosing the left and right pairs of PV. Power settings of 25–30 and 30–35 W were used on the posterior and anterior wall, respectively. Success was defined by achieving complete electrophysiological isolation of the PVs (entrance and exit block) with a circular mapping catheter. At Hospital de Santa Cruz, all patients underwent conventional manually guided ablation, while at Hospital da Luz a Niobe II, magnetic navigation system (Stereotaxis® Inc., St. Louis, MO, USA) was used. Whenever typical atrial flutter (AFL) was documented/observed, additional cavotricuspid isthmus (CTI) ablation was performed. Bidirectional conduction block (endpoint) was assessed by pacing on both sides of the line and recording double potentials separated by > 100 ms throughout its entirety. If required, a direct current (DC) cardioversion was performed at the end of the procedure.

Oral anticoagulation was resumed 6 h after the ablation, maintained for 6 months, and then withdrawn/continued according to CHADS2 score. Patients on vitamin K antagonists with sub-therapeutic international normalized ratio (INR) were kept on subcutaneous enoxaparin 1 mg/kg every 12 h until an adequate INR was achieved. As a general rule, class I/III AAD was maintained in all patients for the first 3 months after the procedure and then withdrawn if there was no AF recurrence, according to physician and patient preferences. A proton pump inhibitor was also prescribed for the first month after the ablation.

2.3 Study endpoint and patient follow-up

The study endpoint was recurrence of AF/AFL/atrial tachycardia (AT) after a 3-month blanking period. Clinical AF was defined as the presence of symptoms likely due to AF episodes. Documented AF was defined by the presence of at least one episode of AF lasting more than 30 s in any ECG or 24-h Holter monitoring.

The follow-up protocol comprised outpatient visits with 12-lead ECG on the 1st, 3rd, 6th, and 12th months post-ablation and 24-h Holter monitoring on the 3rd and 12th months. Through the 2nd to the 5th year post-ablation, the follow-up protocol consistently encompassed a yearly outpatient visit with ECG and 24-h Holter evaluation, after which Holter monitoring was no longer routinely/systematically performed. Patients were also encouraged to contact the department if they experienced symptoms of AF recurrence, after which they were clinically assessed and underwent 12-lead ECG and 24-h Holter monitoring; if the latter could not be performed, clinical relapse was still assumed in order to reduce the risk of underestimating AF recurrence. Whenever clinical records were insufficient, a structured telephonic interview was conducted. Patients who were kept on AAD after the 3rd month of follow-up were not regarded as failed ablation.

2.4 Statistical analysis

Normally and non-normally distributed variables were expressed as mean ± standard deviation and median (interquartile range (IQR)), respectively. To assess differences between groups, independent sample t test and Fisher’s exact test were used for continuous and categorical variables, respectively. Pearson’s correlation with a two-tailed test of significance was used to evaluate LA volume estimation by different methods. Independent predictors of AF relapse were identified by Cox regression analysis. Age, gender, type of AF (paroxysmal vs non-paroxysmal), BMI, hypertension, diabetes, hypercholesterolemia, cigarette smoking, known obstructive coronary heart disease, LV systolic dysfunction, LA volume indexed to body surface area (BSA), CHA2DS2-VaSC, and ATLAS scores were selected for univariate Cox proportional hazards analysis. Variables with a P value ≤ 0.10 were then entered simultaneously in a multivariate Cox regression model and considered to be statistically significant if P value < 0.05—survival curves were plotted according to model results. Multicollinearity was excluded by assessing Pearson’s correlation coefficient between pairs of continuous variables (all < 0.60). To prevent this effect, CHA2DS2-VaSC and ATLAS scores were only assessed in univariate analysis. Kaplan-Meier curves were plotted to evaluate the cumulative incidence of AF relapse, with statistical significance assessed using the long-rank test. Statistical analysis was performed using Statistical Package for Social Sciences (SPSS) version 24.0 (SPSS Inc., Chicago, IL) for Mac OS. Statistical significance was set at a P value < 0.05 (two-sided).

3 Results

The study population (n = 253) had a median age of 55 (IQR 48–63) years and was mostly composed of males (80%) and individuals with paroxysmal AF (64%)—baseline characteristics are summarized in Table 1.

Table 1 Baseline patient characteristics for the whole cohort and according to AAD strategy after the 3-month blanking period
Full size table

Bidirectional PVI was achieved in all patients. At the end of the procedure, 72 patients (28%) underwent DC cardioversion.

During a median follow-up of 5 (IQR 2–9) years, a total of 144 patients (57%) experienced AF recurrence (n = 134) or AT/AFL (n = 10)—annual relapse rate of 10%/year (Fig. 1). Patients who relapsed had a median follow-up of 2 years (IQR 1–4) vs 9 (IQR 9–10) years for patients that remained free from AF. The mean AF-free survival time (amount of time after which 50% of the patients relapsed) was 6.1 ± 2.6 years. Most relapses were established by both symptoms and electrocardiographic documentation (63%), with clinical relapse (28%) and documented silent AF (9%) accounting for the remaining cases. The median follow-up of patients who did not relapse was 9 (IQR 8–10) years.

Fig. 1
figure 1

AF-free survival after a first PVI procedure

Full size image

Using Cox regression multivariate analysis, three characteristics were independently associated with an increased risk of AF relapse in the whole cohort: female sex (adjusted HR 1.526, 95% CI 1.037–2.246, P = 0.032), non-paroxysmal AF (adjusted HR 1.410, 95% CI 1.000–1.987, P = 0.050), and LA volume indexed to BSA (adjusted HR 1.012, 95% CI 1.003–1.021, P = 0.008) (Table 2).

Table 2 Predictors of time to AF recurrence in univariate and multivariate Cox regression analysis
Full size table

During the first year of follow-up, 139 patients (55%) were kept on AAD therapy. Most were prescribed amiodarone (55%), followed by sotalol (19%), flecainide (16%), and propafenone (10%). There were no statistically significant differences between both groups’ baseline characteristics and follow-up period (Table 1). At the 1-year follow-up assessment, overall PVI success rate of patients under AAD was 86 vs 76% with no AAD (P < 0.001). The annual relapse rates for patients who were kept on AAD vs those withdrawn were 8%/year vs 14%/year (P < 0.001), respectively. Kaplan-Meier AF-free survival curves are shown in Fig. 2. Maintaining AAD therapy was associated with a significant long-term reduction in AF recurrence (HR 0.673, 95% CI 0.509–0.904, P = 0.004) assessed by Cox regression multivariate analysis (Table 2). Among patients who were kept on AAD for more than 1 year after the procedure, less than 12% discontinued therapy at the time of their last follow-up assessment. A sensitivity analysis excluding these patients yielded a subgroup of 126 individuals, in which 54 relapsed (44%), with an annual relapse rate of 8%/year—Kaplan-Meier AF-free survival curves are shown in Supplementary Material Fig. II. In this subgroup of patients, AAD remained independently associated with reduced long-term AF relapse (HR 0.698 95% CI 0.521–0.952, P = 0.008).

Fig. 2
figure 2

Kaplan-Meier AF-free survival curves after a first PVI procedure, for patients who were kept vs withdrawn AAD therapy at the end of the blanking period

Full size image

A total of 172 patients underwent PVI using conventional manual navigation vs 81 using a remote magnetic system. Groups based on the type of navigation system used differed significantly on baseline characteristics (indexed LA volume, age, and ATLAS score). Ninety-two patients (54%) relapsed in the manual navigation group vs 46 (57%) in the remote magnetic one (P = 0.685), ascertaining to annual relapse rates of 9.5%/year vs 10.5%year, respectively (P = 0.412) (Supplementary Material Fig. III). Remote magnetic navigation was associated with longer radiofrequency application duration (P < 0.001) and total procedure time (P < 0.001). However, fluoroscopy times were significantly shorter when compared to manual navigation (P < 0.001) (Supplementary Material Table I). Major complications were very rare and similar between both groups (Supplementary Material Table I). Subgroup analysis was consistent with previous results.

4 Discussion

The main findings in our study were as follows: (1) the mean AF-free survival time after a single PVI procedure was 6.1 ± 2.6 years; (2) indexed LA volume, non-paroxysmal AF, and female sex were identified as independent predictors of AF relapse; and (3) maintaining AAD therapy throughout the first year post-ablation was independently associated with longer AF-free survival.

4.1 Single procedure long-term success

In this study, we sought to assess the impact of a single PVI procedure in long-term freedom from AF. The current body of evidence suggests that repeat procedures improve long-term success rates to some degree [9]. However, these are commonly associated with the adoption of supplementary strategies such as performing additional lines or substrate modification, hampering analyses on the effectiveness of isolated PVI. Thus, although reporting the total number of procedures may be clinically useful, its analysis is always affected by how earnestly sinus rhythm is pursued by both patients and physicians.

In our cohort, approximately 80% of the patients remained free from AF at the 1-year follow-up. The observed high success rate was likely due to patients’ baseline characteristics—relatively young population, mostly composed of patients with preserved EF and paroxysmal AF—and the fact that maintaining AAD after the blanking period (~ 55% of the individuals) was not regarded as failed ablation. Indeed, this is corroborated by a systematic review reporting a pooled single-procedure overall success rate of ~ 64% (95% CI 58–73%) at 12 months [9] but in which heterogeneity between the included studies was very high, mostly owing to differences in population characteristics, the use of antiarrhythmic drug after ablation and the definitions of procedural success or failure, among others. Another literature systematic review showed that single-procedure off AAD had a 57% (95% CI 50–64%) success rate, which improved to 72% for patients who were kept on AAD [10]. These results are comparable to those reported in our study. Ouyang et al. also reported similar long-term results after a single PVI procedure in paroxysmal AF patients with 53.4% of the population relapsing over a median 4.8 (IQR 0.3–5.5) years of follow-up period [10].

In our cohort, 31% of all relapses occurred during the first year of follow-up—representing the highest rate of recurrence—after which a mostly steady annual relapse rate of 7%/year was noted. This is consistent with previous multicenter studies reporting long-term outcomes after PVI [15, 16] and shows that even though the first 12 months after PVI represent a critical risk period for AF relapses, the following years are equally important. These results highlight the need to improve upon current follow-up strategies, while continuously promoting lifestyle modifications aimed at reducing systemic/cardiac inflammation and cardiovascular risk, therefore helping to prevent AF relapse and disease progression [9, 17, 18].

4.2 Independent predictors of relapse

Three independent predictors of relapse were identified in this selected population of patients: LA volume indexed to BSA, female sex, and non-paroxysmal AF. Identification of predictors of relapse remains controversial, since studies evaluating clinical outcomes after PVI are markedly heterogeneous with respect to their methods, enrolled patient baseline characteristics, and procedural features [9, 17]. Nevertheless, a growing body of evidence ascertains that certain predictors are associated with different types of early, late, and very late AF recurrence [19]. Similarly to what is frequently reported in the literature [9, 11], patients with a dilated LA and non-paroxysmal AF showed a significantly increased risk of relapse in our study. These characteristics remained associated with both types of long and very long term relapse, supporting previous findings that linked these attributes to the underlying substrate, with an advanced degree of adverse remodeling [20, 21].

It is well established that gender affects how AF is managed [14]. DC cardioversion is less frequently used in women [22], who are also less commonly prescribed class IC/III AAD and less often/later referred for catheter ablation [22,23,24]. However, outcomes after PVI are far less consensual, and despite a large multicenter trial showing that female sex was an independent predictor of relapse after PVI [25], smaller prospective studies reported no differences based on sex [24]. In our cohort, there were no significant differences between the two genders regarding age, cardiovascular risk factors, LA volume, and type of AF. A likely higher incidence of non-PV triggers and gender-related anatomical differences might explain women’s increased risk of AF relapse after catheter ablation.

4.3 Antiarrhythmic drug therapy after PVI

The continued use of AAD after the blanking period and whether this should be regarded as failed ablation persisted as a focus of dispute for more than a decade. The non-uniformity between studies rendered the comparison of outcomes nearly unfeasible [9]. In previous works from our group, we assumed that keeping a patient on AAD should not be classified as relapse, based on the fact that these drugs were not shown to be superior to other agents (e.g., beta-blocker) in maintaining sinus rhythm after PVI [26, 27]. However, the case for maintaining AAD after the 3-month blanking period post-PVI was made significantly stronger after the POWDER-AF trial [28] results were made available. This multicenter randomized controlled trial showed that continued AAD following ablation significantly reduced AF recurrence (2.7 vs 21.9% in the off AAD group, P < 0.0001) and the rate of repeat ablations (1.3 vs 17.1%, respectively; OR = 0.06, 95% CI: 0.001–0.460). Therefore, our results are consistent with the POWDER-AF trial (despite different relapse rates), suggesting that continuing AAD after the blanking period is a protective factor of AF recurrence. Further emphasizing this effect, the EAST-AF trial has also shown that patients who were kept on AAD during the blanking period had a reduced incidence of reccurent atrial tachyarrhythmias during the treatment period, despite similar outcomes at a later phase [29]. These results are in line with the growing body of evidence that the clinical benefit of AAD is lost when therapy is discontinued.

The results comparing the type of navigation used were in line with previous reports from our group [30].

4.4 Limitations

The main limitation of our study is the likely underestimation of AF relapse, specifically the asymptomatic type (silent AF). The follow-up protocol was robust at the time of enrollment (2006–2008) but having continuous ECG-monitoring would certainly improve detection. Nevertheless, since the main goal of AF ablation is to improve symptoms, we believe our pragmatic protocol evaluates the essential benefit of the procedure.

Several characteristics that are currently known to be associated with AF relapse have not been systematically assessed, such as AF burden, obstructive sleep apnea, or LA fibrosis, preventing their inclusion in the Cox regression models. Also, patients with non-paroxysmal AF are likely underrepresented in our cohort.

Extrapolating our findings to other regions such as the US is somewhat limited by our population younger age. Although the median age in our cohort is in line with other reports from approximately the same period these procedures were performed [9] (2007–2008), the median age of patients undergoing PVI in the US was usually higher [31].

The fact that the different technologies used 10 years ago have evolved considerably should also be regarded as a potential limitation.

Finally, this registry was not powered to compare safety outcomes and the decision to maintain AAD therapy after the blanking period not randomized. Additionally, even though there were no significant differences between the two subgroups and adjustments for potentially confounding variables were performed, there still might be significant differences in unrecorded variables.

Despite these limitations, we believe our findings provide valuable information on the very long-term outcomes of patients undergoing a first PVI (without additional lines or substrate modification), as well as clinical management (AAD strategy) of patients after the procedure.

5 Conclusion

Roughly half of the patients remain free from atrial fibrillation 5 years after a single pulmonary vein isolation procedure. Indexed left atrial volume, female sex, and non-paroxysmal AF independently predicted recurrence, whereas continuing antiarrhythmic drug therapy after the 3-month blanking period was associated with longer AF-free survival.