The protocols of this specific study were approved by the Institutional Review Board of Korea University Medicine Anam Hospital (approval number: 2020AN0165). Written informed consent form was waived due to the retrospective nature of this study. We followed the ethical guidelines of the 2013 Declaration of Helsinki in all study protocols.

We screened all AF patients who underwent de novo RFCA in Korea University Medicine Anam Hospital from June 1998 to April 2019. Patient who underwent a voltage mapping were finally enrolled in this study.

Multi-polar catheters were positioned at high right atrium, low right atrium, distal coronary sinus, and right ventricle. Double trans-septal punctures were performed with 2 SL1 sheaths. From February 2019, deflectable sheaths were available and used based on the operator’s discretion. Three dimensional electroanatomic mapping was performed with NavX (St. Jude Medical, St. Paul, MN, USA) system.

Elimination of pulmonary vein (PV) and extra-PV triggers was the end-point of the procedure in paroxysmal AF patients. However, additional substrate modification to achieve non-inducibility was also performed based on the operator’s discretion. For substrate modification in non-paroxysmal AF, AF was induced after completion of PV isolation. If sustained AF (AF lasting more than 5 minutes) was induced, additional substrate modifications including linear ablation and complex fractionated atrial electrogram guided ablation were performed. During the study period, LVZ-guided ablation was not performed in our institution. The precise RFCA protocols of our institution in AF patients are reported in our previous studies.10)16)17)

After double trans-septal punctures and insertion of ablation and mapping catheters, direct-current cardioversion was performed if the rhythm status was AF. Three-dimensional geometry and voltage mapping were performed under paced rhythm unless incessant triggers were observed. Since sinus rate differs among patients, we paced high right atrium with a pacing cycle length of 600 msec in all patients during voltage mapping. The AFocus II circular mapping catheter was used to construct 3-dimensional geometry and acquire bipolar voltage of each point in LA. During electroanatomic mapping, LA surface area (mm²), LVZ area (mm²), LVZ percentage ([LVZ area/LA surface area]×100) were measured. The area beyond PV and LA appendage (LAA) ostium were not included in the measurement.

Prior to RFCA, transthoracic echocardiography (TTE) and transesophageal echocardiography (TEE) were performed. Diameter of LA was measured in parasternal long-axis view in TTE evaluation. Other parameters such as ejection fraction of left ventricle (LV), mitral valve inflow velocity (E), mitral annular tissue velocity (e′), and valve functions were also measured. During TEE evaluation, emptying, filling, and the average flow velocity of the LAA were measured. Spontaneous echo-contrast (SEC) and thrombus in LA or LAA were carefully evaluated with multiple views (high esophageal 0º, 45º, 60º, and 120º views) during TEE evaluation. The grade of SEC was divided into 4 stages: very mild (minimal echogenicity, only detectable transiently, or increasing gain setting required for the detection); mild (detectable without increasing gain setting); moderate (dense, swirling echogenic material, echogenic signal is dense in LAA compared to LA); or severe (dense, swirling echogenic material, echogenic signal is equivocal in LAA and LA). Dense SEC was defined as moderate or severe SEC.

Atrial tachyarrhythmia referred to both AF and atrial tachycardia (AT). Any atrial tachyarrhythmia lasting for more than 30 seconds or documented in 12-lead surface electrocardiogram (10 seconds) 90 days after RFCA was defined as late recurrence (LR). AF was classified as paroxysmal if AF episodes did not last for more than 7 days. Non-paroxysmal AF was defined as AF lasting for more than 7 days or if it required direct-current cardioversion to terminate. LVZ was defined as area in LA in which bipolar voltage range ≤0.5 mV. Substrate modification was defined as any additional ablation lesion other than PV isolation or cavotricuspid isthmus ablation.

Patients were discharged 2 days after RFCA. All patients were recommended to come to our outpatient clinic 2 weeks after discharge to check rhythm status and symptoms. Three months after RFCA, all patients underwent Holter evaluation for 24 hours. Additional Holter monitoring was performed 6, 9, and 12 months post-RFCA. Patients were advised to take portable event recorders whenever they experience symptoms suggestive of recurrence which was not detected by Holter monitoring. Symptom-based standard 12-lead electrocardiography, whether from our or another hospital was also an important part of recurrence detection process. The freedom from LR was calculated for 5 years of follow-up.

Continuous variables are described as mean ± standard deviation and unpaired t-test was used for comparison. Analysis of covariance with post hoc Bonferroni correction was used to compare means of multiple groups. Categorical variables are expressed as percentile value and χ² or Fisher's exact test was used as appropriate. Freedom from LR was depicted using Kaplan-Meier survival curve analysis, and a log-rank test was applied to compare the difference between groups. Multivariable Cox regression analysis was performed to compare freedom from LR while controlling the influence of covariates. Variables that was statistically different according to LVZ area group were selected to be included in the multivariable model. Hazard ratio and its 95% confidence interval (CI) were calculated for each independent variable. Known risk factors for recurrence after RFCA in AF patients were also included in the model.18) Variation inflation factor was calculated to evaluate multicollinearity among covariates included in the multivariate model. The area under curve (AUC) of LA diameter and LVZ to predict LR was calculated by receiver operating characteristic curve analysis. The AUC of LA diameter and LVZ was compared by statistical method reported by Hanley and McNeil.19) Statistical significance was defined as p value <0.05 on a 2-tailed test. All statistical analyses were performed using IBM SPSS Statistics software, Version 24.0 (IBM, Armonk, NY, USA).

Figure 1 summarizes the flow of this study. From June 1998 to April 2019, a total of 3,120 patients underwent de novo RFCA in Korea University Medicine Anam Hospital. Male patients consisted 78.9% of the entire cohort. Mean age and CHA₂DS₂-VASc score were 55.74±10.96 and 1.27±1.26, respectively. Non-paroxysmal AF was diagnosed in 40.9% of patients. Detailed baseline characteristics of our RFCA cohort are summarized in our prior reports.10)20)

During electroanatomic mapping, 537 patients underwent LVZ mapping with AFocus II catheter. TTE was performed in all 537 patients and TEE in 532 (99.1%) patients. Baseline clinical, echocardiographic, and electroanatomic characteristics are described in Table 1. In brief, mean age, CHA₂DS₂-VASc score, and LA diameter were 55.93±10.65, 1.30±1.24, and 41.18±6.14, respectively. Non-paroxysmal AF was diagnosed in 211 (39.29%) patients. Patients were classified into quartiles based on LVZ area. Differences in baseline characteristics among these 4 groups are described in Table 2. Patients with high amount of LVZ had significantly higher prevalence of non-paroxysmal AF and dense SEC; greater LA diameter; lower ejection fraction of left ventricle and LAA flow velocity; and were older. The results of the statistical comparison between each subgroup are described in Supplementary Table 1.

The diameter of LA differed significantly according to LVZ area (p<0.001; Figure 2A). Patients with group 1 (lowest LVZ area; 39.30±4.69 mm) and group 2 (38.84±5.17 mm) had significantly small LA diameter compared with group 3 (41.46±5.66 mm) and group 4 (highest LVZ area; 45.15±6.76 mm). No difference was observed between group 1 and group 2 (p>0.999). The flow velocity of LAA declined significantly as LVZ area increased (p<0.001; Figure 2B).

The diameter of LA was significantly larger in patients with highest LVZ percentage (p<0.001; Figure 2C). However, no significant differences were observed among group 1 to 3 in post hoc analysis (Figure 2C). The flow velocity also differed according to LVZ percentage with group 4 (highest LVZ percentage) having lowest LAA flow velocity (p<0.001; Figure 2D).

The risk of LR was significantly influenced by LVZ area quartile with group 4 (highest LVZ area) having the highest risk (p<0.001; Figure 3A). The percentage of LVZ was also associated with the risk of LR (p<0.001; Figure 3B). In multivariable Cox regression analysis, LVZ area as a quartile or 2 groups were significant risk factors for LR after RFCA (Table 3). Age, LA diameter, ejection fraction of left ventricle, E over e′, and LAA flow velocity were all significantly associated with the risk of LR in univariable analysis. However, these variables were not associated with LR in multivariable analysis with only AF type and LVZ being significant risk factors for LR.

The diameter of LA (AUC, 0.592; 95% CI, 0.532–0.653; p=0.002), LA surface area (AUC, 0.593; 95% CI, 0.532–0.654; p=0.002), LVZ area (AUC, 0.676; 95% CI, 0.622–0.729; p<0.001), and LVZ percentage (AUC, 0.671; 95% CI, 0.617–0.724; p<0.001) were all able to predict the occurrence of LR. In AUC comparison analysis, LA diameter and LA surface area had similar predictive values for LR (0.593 vs. 0.592; p=0.987; Figure 4A). In contrast, LVZ area (0.676 vs. 0.592; p=0.011; Figure 4B) and LVZ percentage (0.671 vs. 0.592; p=0.027; Figure 4C) had significantly superior predictive values for LR compared with LA diameter.