This study was performed in 2 centers, Seoul National University Hospital and Korea University Guro Hospital, using an AF ablation cohort comprising patients with available genomic DNA data. Consecutive patients who were admitted and underwent de novo or repeat radiofrequency catheter ablation (RFCA) for symptomatic paroxysmal or persistent AF between June 2008 and March 2015 were enrolled. A detailed medical and personal history of each participant was obtained at the time of admission. Transesophageal and transthoracic echocardiography were performed prior to catheter ablation to exclude the presence of atrial thrombi and to measure the cardiac chamber size, wall thickness, and left ventricular systolic function with standard methods. After the index catheter ablation procedure, the prescription of antiarrhythmic drugs during the blanking period was left to the attending clinician's judgement. Follow-up information was obtained from regular outpatient visits at 1, 3, 6, and 12 months and every 3 to 6 months thereafter, as clinically indicated. Electrocardiograms were performed at every visit, and 24-hour Holter monitoring was performed at 3 and 12 months after the ablation. Supplementary 24-hour Holter monitoring was obtained when recurrence was suspected on the basis of the patient's symptoms.

The primary endpoint was the recurrence of atrial tachyarrhythmia after AF ablation. Recurrence was defined as any documented episode of AF, atrial flutter (AFL), or atrial tachycardia (AT) lasting more than 30 seconds after a 3-month blanking period.19) The study adhered to the tenets of the Declaration of Helsinki and was exempt from review by the Institutional Review Board of Seoul National University Hospital Biomedical Research Institute (No. 1408-106-605). All subjects provided written informed consent.

Ablation was guided by 3-dimensional electroanatomical mapping using CARTO (CARTO XP or CARTO-3; Biosense Webster Inc., Diamond Bar, CA, USA) or EnSite (EnSite NavX Classic or EnSite Velocity; Abbott Laboratories, North Chicago, IL, USA) mapping system. A duo-decapolar Lasso circular mapping catheter was used to guide and map the PV. Ablation was performed using open irrigation catheters (Celsius or Navistar Thermocool SF; Biosense Webster Inc., Cool Flex; Abbott Laboratories). All patients underwent circumferential PV isolation. Exit and entrance blocks were confirmed after PV isolation. For patients with persistent AF, additional ablation was performed at the roof line, posterior inferior line, anterior line, mitral isthmus line, cavotricuspid line, or regions of complex fractionated electrograms, at the operator's discretion.

Through a comprehensive screening of previous reports, we selected 20 single nucleotide polymorphisms (SNPs) from genome-wide association studies in which the robust association between SNPs and incident AF were identified (p<5×10⁻⁸).12)14)15)16)17) The selected SNPs are listed in Supplementary Table 1, along with the risk allele frequencies in our cohort and published effect sizes for incident AF. All SNPs have a reported minor allele frequency of >0.01.

Genomic DNA extraction was performed by standard methods with whole blood samples collected during admission for index catheter ablation. The selected 20 SNPs were genotyped using validated TaqMan SNP genotyping assays (Applied Biosystems, Foster City, CA, USA) on the ABI PRISM 7900HT Real-time polymerase chain reaction system according to the supplier's recommendations, and allelic discrimination was calculated with SDS v2.3 software (Applied Biosystems) using the same system. The laboratory personnel were blinded to the subjects' clinical characteristics and outcomes of the ablation procedure. All SNPs had a call rate of >99%, and agreement of genotype frequencies with Hardy-Weinberg equilibrium expectations was tested by using a χ² goodness-of-fit test.

To estimate genetic risk of AF recurrence using a minimal set of genetic variants, candidate SNPs were selected before the GRS construction. The 20 SNPs were separately examined for independent cross-sectional association with AF recurrence using a logistic regression analysis, and SNPs showing at least borderline significant associations (p<0.1) with AF recurrence were included in the final GRS model. Because the associations between each AF susceptibility-associated SNP and clinical response to catheter ablation were not previously reported in most cases, the magnitude of genetic effect size of each SNP could not be estimated. Therefore, we applied an additive unweighted model, and the total number of risk alleles each subject carries was summed to calculate the GRS. Previous research has shown that unweighted GRS model gives unbiased prediction of the associations, when there is a lack of relevant data on the individual effects of each genetic variant.20) Subjects with missing genotype data of the target SNPs included in the final GRS model were excluded from the analysis.

Baseline characteristics are presented as mean±standard deviation for continuous variables and as number and percent for categorical variables. A Cox multivariable proportional hazard model was used to identify the association between the GRS and risk of AF recurrence after catheter ablation. Baseline characteristics including age, sex, body mass index (BMI), persistent (vs. paroxysmal) AF, hypertension, diabetes mellitus (DM), heart failure, previous history of stroke, left atrium (LA) diameter, and left ventricular ejection fraction (LVEF) were used to adjust the endpoint. Among the confounders included in the multivariate analysis, age, LA diameter, and LVEF were used as continuous variables, while others were applied as categorical variables. We divided the subjects into 3 subgroups by GRS risk, and the differences in the cumulative AF recurrence rate between the subgroups were calculated by the Kaplan-Meier method and compared by the log-rank test. Comparisons of baseline characteristics of each subgroup were made with independent-samples t-tests for continuous variables and χ² tests for categorical variables. The risks of AF recurrence in the intermediate and high GRS groups were compared with the low GRS group using a Cox multivariable proportional hazard model, adjusted for the same risk factors. All analyses were performed using SPSS Statistics 20.0 (SPSS Inc., Chicago, IL, USA).

A total of 746 subjects were included. The characteristics of the study population are summarized in Table 1. The median follow-up time was 684 days (interquartile range, 324–1,205 days). During the follow-up period, 168 subjects (22.5%) experienced recurrence of atrial tachyarrhythmia. AF accounted for 57.1% of all recurrence events, and AFL or AT accounted for the remainder. The 6- and 12-month recurrence rates were 8.2% and 14.7%, respectively (Table 1). The overall recurrence patterns throughout the study period are presented in Supplementary Figure 1.

All of the 20 selected SNPs were in accordance with Hardy-Weinberg expectations (p>0.001). We evaluated each SNP separately to assess the cross-sectional association between each of the 20 SNPs and recurrence of atrial tachyarrhythmia, and to identify an essential set of SNPs for joint analysis by a GRS. We assumed that each copy of the risk alleles adds an equal risk effect, and an unweighted additive model was applied in the logistic regression analysis.

Table 2 presents the associations of the individual genetic variants with recurrent AF after catheter ablation. Overall, 6/20 SNPs showed at least borderline association (p<0.1) with recurrence: 5 SNPs at the 4q25 locus (rs1448818, rs6817105, rs2200733, rs6843082, and rs6838973) and 1 SNP at the 16q22 locus (rs2106261). Although modest trends were noted, none of these selected SNPs remained significantly associated with recurrence in the multivariable Cox proportional hazard model adjusted for traditional risk factors (Supplementary Table 2). Of these 6 SNPs, rs6817105 was excluded from the final GRS model because of its strong linkage disequilibrium (r²=1.0) with rs2200733 (Supplementary Table 3), which has been reported as an independent predictor of treatment response after catheter ablation of AF in different populations.16)17)

The primary endpoint of the study was AF recurrence, defined as any episode of atrial arrhythmia including AF, AT, or AFL.19) To assess the cumulative effects of genetic variants on AF recurrence, we performed a joint analysis by generating a GRS with the 5 selected SNPs. As we adopted an unweighted additive model, the GRS was calculated by summing the number of risk alleles for each of the 5 selected SNPs, which resulted in a score ranging from 0 to 10. The overall distribution of the GRS is shown in Figure 1. Although the distribution of the GRS between those who experienced recurrence during follow-up and those who remained in sinus rhythm tended to overlap, the mean values of each group significantly differed. The mean GRSs were 6.3±1.7 and 5.7±1.8 in the AF-recurrence and no-recurrence groups, respectively (p<0.001).

Cox regression analysis demonstrated a significant association between the GRS and AF recurrence after adjusting for baseline characteristics, including age, sex, BMI, type of AF, hypertension, DM, heart failure, previous history of stroke, LA diameter, and LVEF. The addition of 1 risk allele was associated with a 13% increased risk of recurrence (hazard ratio [HR], 1.13; 95% confidence interval [CI], 1.03–1.24; p=0.007) (Table 3). To illustrate this association further, the subjects were partitioned into 3 groups according to the GRS: the low-risk reference group (GRS 0–3), intermediate-risk group (GRS 4–6), and high-risk group (GRS 7–10). The epidemiologic and echocardiographic data of each group are presented in Supplementary Table 4. There were statistically significant differences in the type of AF and LA size among the groups, with subjects with higher GRS tending to have a higher prevalence of persistent AF and larger LA. Figure 2 shows Kaplan-Meier curves of the proportions of subjects remaining in sinus rhythm in the 3 groups, which revealed that there was a significant trend toward an increased risk of AF recurrence in the group with higher GRS (log-rank test; p=0.003). Compared with the low-risk reference group (GRS 0–3), subjects with intermediate-risk (GRS 4–6) tended to be associated with a higher risk of recurrence, with an adjusted HR of 2.00 (95% CI, 0.99–4.04; p=0.055). Subjects at high-risk (GRS 7–10) showed further increased risk of recurrence, with an adjusted HR of 2.66 (95% CI, 1.32–5.37; p=0.007) (Figure 3). When we put a narrow definition on the AF recurrence by censoring AFL or AT events at the time of the first episode, the GRS maintained the performance of the recurrence prediction after AF ablation: Cox regression analysis demonstrated a significant association between the GRS and AF recurrence after multivariable adjustment (HR, 1.49; 95% CI, 1.07–2.08; p=0.019), and each GRS subgroup showed different recur profile (log-rank test; p=0.014).

In the current study, we enrolled consecutive patients who underwent catheter ablation for symptomatic AF, and study subjects constitute a heterogeneous group with variable epidemiologic and procedural characteristics. In the procedural aspect, both de novo (696/746, 93.3%) and redo (50/746, 6.7%) ablation procedures were performed, which have different procedural strategies and clinical outcomes. We performed a subgroup analysis including only subjects underwent de novo AF catheter ablation with the same outcome variables. The results were substantially similar to the analysis of overall population. After adjusting for the clinical risk factors, the GRS was significantly associated with AF recurrence (HR, 1.15; 95% CI, 1.04–1.27; Supplementary Table 5), and graded risk of recurrence was demonstrated according to the GRS tertiles (Supplementary Figure 2). On the other hand, statistically significant predictor of AF recurrence was not found in the patients who underwent redo ablation.