A single-center, retrospective cohort method was performed in this study. Patients with PAF admitted to the Changzhou Hospital of Traditional Chinese Medicine from January 2021 to August 2023 and scheduled to undergo RFA were enrolled in this research. Inclusion criteria were as follows: (1) older than 18  years old; (2) diagnosis of PAF (defined as a sustained episode lasting <7 days or chemical and/or electrical cardioversion <7 days); (3) sinus rhythm when mapping; (4) patients who had received PVI alone; (5) complete basic information, procedure, and follow-up data. Exclusion criteria included: (1) severe organic heart disease (such as severe mitral regurgitation or stenosis, severe aortic regurgitation or stenosis, hypertrophic obstructive cardiomyopathy, etc.); (2) history of previous cardiac surgery; (3) patients with incomplete clinical and follow-up data. This study adhered to the Helsinki Declaration and received approval from the Scientific Ethics Committee of Changzhou Hospital of Traditional Chinese Medicine (2023-LL-017l); it was already registered in the Chinese Clinical Trial Registry (ChiCTR2300079137).

Catheter ablation was carried out in the CARTO system. All procedures were performed without interruption of anticoagulant therapy and off antiarrhythmic medications. The high-density multielectrode mapping catheter [PentaRay (Biosense Webster, Inc.)] and irrigated contact force sensing ablation catheter [Thermocool SmartTouch ST or STSF (Biosense Webster, Inc.)] were introduced into the left atrium with transseptal puncture.

PentaRay reconstructed the electroanatomical map of the left atrium, including the main part of the left atrium, the left atrial appendage (LAA), and the pulmonary veins (PVs) after excluding the inspiratory phase through fast anatomical mapping (FAM). Then, ST or STSF examined the true boundary of the map further through contact force. Thus, a more realistic LA model could be revealed.

Ablation was performed with ST or STSF at a power of 35 W. The discharge time was determined according to the AI at each tag, with an intertag distance ≤6 mm. A target AI of 480–530 for the roof, anterior, and ridge; 420–450 for the inferior; and 360–380 for the posterior were recommended. PVI was defined by PV entrance and exit block.

Clinical baseline indicators of patients were collected, including gender, age, body mass index (BMI), duration of AF, CHA₂DS₂-VASc score, and medical history [including heart failure (HF), hypertension, diabetes mellitus (DM), and stroke or transient ischemic attack (TIA)].

Laboratory indicators [including urea nitrogen, creatinine, and N-terminal precursor protein brain natriuretic peptide (NT-proBNP)], and echocardiographic indicators [including the left atrial diameter (LAD), left ventricular end-diastolic diameter (LVEDD), left ventricular end-systolic diameter (LVESD), and left ventricular ejection fraction (LVEF)] were collected.

Anatomical and electrical parameters were measured in the 3D electroanatomical map of the CARTO system. All parameters were obtained from the maps before ablation.

Because of the embryonic homology between the PW and the PVs, the area of the PV ostia was also measured in this study. A new map was saved (Map 1) after obtaining the initial map. In Map 1, we drew a short line along the axial direction of PVs from the superior, inferior, anterior, and posterior orientation to assist us in the next step (Figure 1). Then, we utilized the function of “clipping pane” to make a vertical section at the PV ostia [the right superior pulmonary vein (RSPV), the right inferior pulmonary vein (RIPV), the left superior pulmonary vein (LSPV), and the left inferior pulmonary vein (LIPV), respectively]. The measurement would be merged in the common pulmonary vein, and the accessory pulmonary vein was also measured if it resembled the main bifurcation. Subsequently, the diameters of the long axis (a) and the short axis (b) of each PV were measured separately (Figure 1). According to the calculation formula ellipse area = π × (a/2) × (b/2), the PV ostia area was obtained. The total area of the right PV (RPV) and left PV (LPV) was added up by each bifurcation. All parameters were measured independently by two researchers and an interobserver agreement was reached.

A map was saved again (Map 2) after completing this step. In Map 2, we drew another short line at the roof and the bottom of the PW. The inferior line was defined as a line located at the lowest margin between the RIPV and the LIPV in the posterior view (Figure 2a). The superior line was defined as a line that connected the middle of the LSPV and RSPV ostia in the superior view (SUP) (Figure 2b). The PW surface area (PWSA) was composed of the area delineated by the superior and inferior line and 0.5–1 cm outside the PV ostia. Then, the function of “area measurement” was used to measure and calculate the area (Figure 3).

The next step was to copy activation mapping points in the PW area to another map (Map 3) (Figure 4). In Map 3, after switching the bipolar map, we sorted all copied points by voltage (Figure 5). A very low-voltage area (VLVA) in the electrogram was defined as an amplitude of ≤0.1 mV. We subsequently counted the points with voltage ≤0.1 mV and calculated the percentage of the VLVA (PVLVA) in the LAPW (Figure 5)

In this study, we divided the distribution of the VLVA in the LAPW into two parts, namely, the connection of PVs and the LAPW (PV ostium in the PW) and the intra-atrial region in the PW (Figures 6a,b).

Procedural success was defined as freedom from recurrent atrial arrhythmias lasting longer than 30 s after an initial 3-month blanking period. Recurrent arrhythmias were classified as being AF, atrial tachycardia, or both. Patients underwent a clinical review and 24 h Holter monitoring at 3, 6, and 12 months.

Statistical analysis was performed using SPSS Statistics Software version 26 (IBM, Inc.). Measurement data that followed a normal distribution pattern are expressed as mean ± standard deviation (x ± s). In cases where these data had a mildly skewed distribution, the median was used. Either an independent sample t-test or the Kruskal–Wallis test was performed in intergroup comparisons. Counting data were presented as a percentage, and intergroup comparisons were conducted using the chi-square test or Fisher's precision probability test. A univariate logistic regression analysis assessed the association of different factors and AF recurrence. After collinearity analysis (tolerance <0.1 or variance inflation factor >10 meant multicollinearity between variables), a multivariate logistic regression analysis was employed to evaluate which factor was an independent risk factor for AF recurrence. A p-value <0.05 was considered statistically significant. If a covariate that changed the estimates of AF recurrence had a significant association (p < 0.10) in the univariate logistic regression analysis, it would be included as a potential confounding factor in the multivariate logistic regression analysis. A receiver operating characteristic (ROC) curve was used to construct the combined prediction model for AF recurrence. If the area under the curve (AUC) was greater than 0.75, it indicated better predictive ability. A Hosmer–Lemeshow goodness-of-fit test was performed to evaluate the calibration ability of the prediction model, and a p-value >0.05 was considered a good fit of the model.

According to the inclusion and exclusion criteria, a total of 128 patients were enrolled in this study. Of the initial participants, eight were excluded from the final analysis (including six incomplete data and two lost to follow-up) (Figure 7). Consequently, the baseline characteristics of 120 patients are presented in Table 1. According to the occurrence or non-occurrence of recurrent arrhythmias, these patients were divided into Group 1 (no recurrence) and Group 2 (recurrence). Group 1 comprised 49 males and 60 females, averaging 67.66 ± 12.60 years. Group 2 was composed of four males and seven females, averaging 67.18 ± 13.23 years. There was no significant difference between the two groups in terms of gender, age, BMI, duration of AF, CHA₂DS₂-VASc score, history of hypertension, DM and stroke, or TIA, except for a history of HF (p = 0.028). There was also no statistical difference in terms of urea nitrogen, creatinine, NT-proBNP, LAD, LVEDD, LVESD, and LVEF (Table 1).

Procedural characteristics are presented in Table 2. The areas of the PV ostia (RPV and LPV, respectively) and the PWSA and the PVLVA were measured and calculated. The complications were recorded. Among the procedural characteristics, patients in Group 2 were found to have a significantly larger PWSA (p = 0.013) and PVLVA (p < 0.001) (Table 2).

In patients with the VLVA in the LAPW, the proportions of distribution of the VLVA in the PV ostium in the PW, the intra-atrial region in the PW, and the mixed areas (contained both) were 60.0%, 20.0%, and 20.0%, respectively, in Group 1 and 12.5%, 12.5%, and 75.0%, respectively, in Group 2 (p = 0.026) (Table 3).

AF recurrence was used as the dependent variable and baseline and procedural characteristics as independent variables. A univariate logistic regression analysis was performed to identify the potential factors associated with AF recurrence. The result showed that the BMI [odds ratio (OR): 0.839, 95% confidence interval (CI): 0.682–1.032, p = 0.096], HF (OR: 3.831, 95% CI: 1.080–13.585, p = 0.038), PWSA (OR: 1.369, 95% CI: 1.056–1.775, p = 0.018), and PVLVA (OR: 1.064, 95% CI: 1.025–1.103, p = 0.001) were potential factors related to AF recurrence (p < 0.10) (Table 4).

Although the p-value of the LAD was >0.10 in the univariate logistic regression analysis, it was a common influencing clinical factor, which was also included in the multivariate logistic regression analysis. The collinearity analysis of the BMI, HF, LAD, PWSA, and PVLVA were performed at first. The results showed that there was no linear relationship between these variables (Table 5). Then, a multivariate logistic regression analysis of potential factors demonstrated that both the PWSA and the PVLVA were independent risk factors for AF recurrence (OR: 1.457, 95% CI: 1.037–2.049, p = 0.030; OR: 1.059, 95% CI: 1.013–1.107, p = 0.012, respectively) (Table 6).

According to the logistic regression analysis of this study, the HF, LAD, PWSA, and PVLVA were included to construct the combined prediction model for AF recurrence and to draw the ROC curve (Figure 8). The equation was Logit (P) = −0.057 + 1.296 × HF − 0.175 × LAD + 0.342 × PWSA + 0.066 × PVLVA. The finding revealed that the AUC of this model was 0.900 (0.827–0.973) (p < 0.001) (Figure 8). The Hosmer–Lemeshow goodness-of-fit test was performed to evaluate the calibration ability of the prediction model (Figure 9). The result suggested that there was no statistical difference between the predicted value of the model and the observed value (χ² = 4.643, p = 0.796).