Consecutive patients with nonvalvular paroxysmal or persistent AF who underwent a single catheter ablation from March 2013 to December 2015 in Tottori University Hospital, Japan were enrolled and followed prospectively for the recurrence of AF after ablation. Paroxysmal AF was defined as self-terminating AF lasting for up to 7 days or AF episodes cardioverted within 7 days, and persistent AF as AF lasting between 7 days and 1 year or AF episodes terminated by cardioversion either with drugs or by direct current cardioversion after 7 days or more [5]. Levels of estimated glomerular filtration rate (eGFR) were calculated with the following equation established by the Japanese Society of Nephrology: eGFR (mL/min/1.73 m²)  = 194 × serum creatinine level^−1.094 × age^−0.287 (× 0.739 if female).[3] The CHADS₂ score comprised congestive heart failure, hypertension, age ≥ 75, diabetes mellitus, and stroke/transient ischemic attack and the CHA₂DS₂-VASc score comprised congestive heart failure, hypertension, age ≥ 75 years, diabetes mellitus, stroke/transient ischemic attack, vascular disease, age 65–74 years, and female sex [13]. The APPLE score comprised age > 65 years, persistent AF, reduced eGFR < 60 ml/min/1.73 m², left atrial diameter (LAD) ≥ 43 mm, and left ventricular ejection fraction (LVEF) < 50%. The APPLE had better prediction of AF recurrence between 3 and 12 months after catheter ablation than CHADS₂ and CHA₂DS₂-VASc scores [17]. Patients with one or more of the following were excluded from candidates for ablation of AF in this study: age less than 20 years or over 85 years; cerebral or myocardial infarction within 6 months of onset; uncontrolled heart failure or left ventricular ejection fraction < 40%; severe chronic obstructive pulmonary disease; contraindications to anticoagulants; thrombus formation within the left atrial appendage; LAD > 60 mm; moderate or severe valvular insufficiency or stenosis; connective tissue disease; and active cancer. It was also planned to exclude patients with hypertrophic/dilated cardiomyopathy or old myocardial infarction, end-stage renal disease or on hemodialysis: five patients with hypertrophic cardiomyopathy and six with end-stage renal disease or on hemodialysis were thus excluded. This study was approved by the Ethics Committee of Faculty of Medicine, Tottori University, and conducted in compliance with the ethical principles of the Declaration of Helsinki. Written informed consent was obtained from all individual participants included in the study.

The protocol for catheter ablation of AF has been reported previously [11]. Briefly, prior to the ablation procedure, all patients received effective oral anticoagulation therapy for at least four weeks and underwent transesophageal echocardiography and enhanced cardiac computed tomography for 3D mapping. Antiarrhythmic drugs were discontinued for more than five times their half-lives. Oral anticoagulants were discontinued before the procedure: direct oral anticoagulants for one day and warfarin for five days, and replaced with unfractionated heparin until 6 h prior to the ablation procedure. Bipolar electrograms were continuously recorded on a digital recording system (Labsystem PRO; Bard Electrophysiology, Lowell, MA, USA). A 20-pole three-sited mapping catheter (BeeAT; Japan Lifeline, Tokyo, Japan) was used to record bipolar electrograms of the superior vena cava, right atrium, and coronary sinus. After transseptal puncture, heparin was continuously infused to maintain an activated clotting time of 300–400 s. Two decapolar circular mapping catheters (Lasso; Biosense Webster, Diamond Bar, CA, USA) were located in the pulmonary veins. A 3.5 mm irrigated-tip ablation catheter (Thermocool; Biosense Webster) was advanced into the LA to achieve bilateral circumferential pulmonary vein isolation with the endpoint of bidirectional conduction block between the LA and pulmonary vein. A point-by-point radiofrequency current was delivered for 30 s with a power of up to 40 watts and a target temperature less than 43°C using a 3D mapping system (CARTO3; Biosense Webster), whereas the power was limited to 20 watts at the site close to the esophagus. After achieving pulmonary vein isolation, a bidirectional conduction blocking line was created at the cavotricuspid isthmus in all patients. The ablation procedure was performed under conscious sedation with continuous monitoring of blood pressure and oxygen saturation.

The primary clinical outcome was recurrent AF lasting over 30 s off antiarrhythmic drugs. Episodes during the blanking period of the first 90 days after the ablation procedure were not considered as recurrences [5]. All patients underwent continuous ECG monitoring for 3 days following the procedure until discharge. Follow-up checks were performed 1, 2, 3, 6, 9, 12, 16, 20, and 24 months after the ablation procedure and comprised physical examination, 12-lead electrocardiogram, and 24-h Holter monitoring. Patients were instructed to self-monitor their pulses daily and report any irregularity of pulse immediately, at which point event monitoring or 14-day Holter monitoring was performed. Oral anticoagulant therapy was continued in all patients with CHA₂DS₂-VASc scores of two or more points after the procedure, but discontinued at 6 months in those whose scores were less than 2 points.

Categorical data are presented as frequency (percentage [%]) and were compared with Fisher’s exact test. Continuous data are presented as mean (standard deviation [SD]) or as median (interquartile range) for skewed distributions. Normality was tested with the Kolmogorov–Smirnov test. Normally distributed continuous variables were compared using the independent Student’s t test and skewed data using the non-parametric Mann–Whitney U test. Analyses were performed after natural logarithmic transformation (ln) of skewed biomarkers. Freedom from AF was reported as crude event rates and by means of a time-to-event analysis using the Kaplan–Meier method. Variables with P value < 0.10 in the group comparison were further evaluated by the Cox proportional hazards regression analyses to determine their associations with the recurrence of AF. The variables with P value < 0.10 in the univariate Cox proportional hazards regression analysis were further evaluated by the multivariate Cox proportional hazards regression analysis. Multiple regression analysis via the forced entry procedure was performed to test correlations of a normally distributed continuous variable with other variables. Statistical analyses were performed using R software (version 3.4.0; R Foundation for Statistical Computing, Vienna, Austria), and P value < 0.05 was considered to denote statistical significance.

In total, 254 patients with AF (mean age of 65.5 [SD =9.3]; 68.9% paroxysmal and 31.1% persistent; all first session) were included in this study (Table 1). All ablation procedures were successful. The median length of follow-up for censored cases was 730 days. AF recurred in 65 (25.6%) patients during follow-up. Ln BNP was significantly greater in patients with the recurrence of AF than those without the recurrence of AF (P = 0.006) and was an only variable which significantly differed between the groups (Table 1). Although statistically not significant, uses of β-blocker and loop diuretics were more frequent and ln sTM were lower in patients with the recurrence of AF than those without the recurrence of AF (P value < 0.10, Table 1).

Because baseline ln BNP was significantly greater in patients with the recurrence of AF than those without the recurrence of AF, the study patients were divided into two groups by the median concentration of BNP (groups with lower and higher levels of BNP) and then a probability of remaining free of the recurrence of AF after ablation procedure with a three-month blanking period was evaluated in the two groups. As a result, patient with the lower level of BNP had a greater probability of remaining free of the recurrence of AF than those with the higher level of BNP (P value by log-rank test = 0.011) as shown in Fig. 1. Two-year recurrence–free survival rates were 0.812 (95% confidence interval [CI] =  0.730–0.871) and 0.672 (95% CI = 0.578–0.749) in patients with the lower and higher levels of BNP, respectively.

To determine associations of baseline variables with the recurrence of AF, the univariate and multivariate Cox proportional hazards regression analyses. In the univariate Cox proportional hazards regression analyses, only ln BNP was significantly associated with the recurrence of AF (unadjusted HR = 1.353, 96% CI = 1.068–1.713, P = 0.012), whereas use of β-blocker, use of loop diuretics, and ln sTM were not significantly associated with the recurrence of AF (unadjusted HR = 1.582, 96% CI = 0.968–2.585, P = 0.067; unadjusted HR = 1.603, 96% CI = 0.873–2.945, P = 0.128; unadjusted HR = 0.447, 96% CI = 0.180–1.107, P = 0.082, respectively) as shown in Table 2. In the multivariate Cox proportional hazards regression analysis, ln BNP remained significantly associated with the recurrence of AF (adjusted HR = 1.286, 96% CI = 1.000–1.655, P = 0.049) when tested with the use of β-blocker and ln sTM, P values of which were < 0.10 in the univariate Cox regression analysis (Table 2).

Because ln BNP remained significantly and independently associated with the recurrence of AF, we performed a multiple regression analysis to test correlations between ln BNP and the key risk variables by the forced entry procedure. In a multivariate regression model controlling simultaneously for rhythm at blood collection, age, sex, type of AF, body mass index (BMI), eGFR, LAD, and LVEF, rhythm at blood collection, age, sex, LAD, and LVEF were significantly associated with the level of ln BNP (Table 3): specifically, AF at blood collection was significantly associated with the substantially increased level of ln BNP (adjusted β = 0.482, 95% CI = 0.213–0.751, P value < 0.001).

We further performed subgroup analyses of risks of the recurrence of AF for ln BNP by dividing patients with the risk variables and the interaction effects of ln BNP with other risk variables, specifically with various combinations of rhythm status at blood collection and other risk variables because rhythm status at blood collection was significantly associated with substantially different levels of ln BNP (Table 3). Consequently, there was a significant interaction of ln BNP with sex and rhythm at blood collection (P for interaction = 0.011) as shown in Table 4: higher ln BNP was significantly associated with an increased risk of the recurrence of AF in a subgroup of male sex with sinus rhythm at blood collection (HR = 1.536, 95% CI = 1.095–2.154, P value = 0.013), but not in the other subgroups.

Many observational studies have reported predictors of the recurrence of AF after catheter ablation including the key risk variables [25]. Persistent AF and LAD, both of which are related to atrial remodeling [11,12], are most consistently recognized as important baseline predictors for the recurrence of AF [25]. A study showed that the APPLE score for the prediction of rhythm outcomes after catheter ablation of AF, which comprises age, type of AF, renal function, LAD, and LVEF, had a better prediction of the recurrence of AF than CHADS₂ and CHA₂DS₂-VASc scores [17]. However, in this study, above-mentioned baseline clinical variables, CHADS₂, CHA₂DS₂-VASc, APPLE scores were not significantly different between patients with and without the recurrence of AF. Mean age of patients enrolled in this study was 65.5 years, whereas those of patients in other studies to identify predictors of late recurrence after AF ablation, to evaluate the values of CHADS₂, CHA₂DS₂-VASc, and the APPLE scores were 57–61 years old [16,17,25]. Furthermore, BMI and LAD of patients in this study were smaller, a period of follow-up longer than those of patients in other studies, and races of patients differed between studies. Therefore, these differences between the current study and other studies may have caused inconsistent results of predictive values of baseline clinical variables and scoring systems for the recurrence of AF.

Inflammation and oxidative stress have been reported as the etiology of AF development [8‐10] and biomarkers of endothelial dysfunction caused by inflammatory, oxidative stress have been reported to be associated with risk of incident AF [20‐24]; however, an association between pre-ablation concentrations of endothelial dysfunction biomarkers and the recurrence of AF has not yet been investigated. In the current study, we found no significant endothelial dysfunction biomarkers associated with risk of the recurrence of AF. It has been reported that the risk associated with endothelial dysfunction for arrhythmia recurrence following catheter ablation was age-dependent and was higher in younger than older patients with AF [23]. Therefore, it is possible that baseline concentrations of endothelial dysfunction biomarkers have predictive values for the recurrence of AF in younger patients than those enrolled in the current study. Further study is needed to confirm this possibility.

Associations between natriuretic peptide and the recurrence of AF have been intensively studied. Meta-analyses have shown that the recurrence of AF was associated with higher pre-ablation concentrations of BNP than non-recurrence although results of the studies were significantly heterogeneous; however, such heterogeneities have not been fully explained [18,19]. In the current study, we found that baseline ln BNP was significantly and independently associated with the recurrence of AF. We also found that AF at blood collection was significantly associated with the substantially increased level of ln BNP. Moreover, the subgroup analyses showed a significant interaction of ln BNP with sex and rhythm at blood collection, which may explain the heterogeneities between studies to test associations between BNP and the recurrence of AF [18,19]. Taken together, in patients with nonvalvular AF, baseline BNP was suggested as a useful biomarker for the risk of late recurrence of AF after ablation; however, there is a need to be careful while using BNP as a biomarker for the recurrence of AF by taking account of the effects of rhythm status at blood collection and sex difference.

Several limitations of our study need to be considered. This study was performed in a single center, which may have resulted in a selection bias: indeed, BMI and LAD of patients in this study were smaller, and age was higher than those of patients in other studies. Asymptomatic episodes of AF may have been missed. We did not evaluate the responses of biomarkers to ablation therapy. We were unable to establish causal relationships between biomarker concentrations and the recurrence of AF because of a prospective observational study. However, our study also had several strengths, including a period of follow-up longer than other studies, deep and systematical evaluation of associations of the risk scores, biomarkers and atrial remodeling with the recurrence of AF, which allowed us to recognize the usefulness of BNP as a predictor of recurrence after ablation treatment of atrial fibrillation and at the same time to recognize the attention at that time.