Predictive value of peak left atrial reservoir strain combined with left atrial appendage flow velocity for atrial fibrillation recurrence after pulsed field ablation

Abstract

Background

The recurrence rate after ablation for atrial fibrillation (AF) remains high, and the predictive performance of existing predictors is limited. Left atrial dysfunction plays a key role in the occurrence and progression of AF. However, studies on predictors of recurrence after pulsed field ablation (PFA) remain relatively limited, particularly comprehensive assessments combining left atrial strain parameters with left atrial appendage flow velocity.

Methods

Clinical data from 246 patients with AF who underwent PFA between June 2021 and June 2024 were retrospectively analyzed. Preoperative peak left atrial reservoir strain (LASr) was measured using two-dimensional speckle-tracking echocardiography (2D-STE), and left atrial appendage flow velocity (LAAV) was measured using transesophageal echocardiography (TEE). The follow-up cutoff date was December 30, 2024, and the median follow-up duration was 17.3 (10.8, 25.6) months. Recurrence of AF, atrial flutter, or atrial tachycardia after the blanking period was recorded. A predefined multivariable Cox proportional hazards model was used to identify independent predictors, and model performance was evaluated using the 18-month time-dependent receiver operating characteristic (ROC) curve, Kaplan-Meier method, and bootstrap internal validation.

Results

During follow-up, 68 patients (27.6%) experienced recurrence of atrial arrhythmia after the blanking period. Both LASr and LAAV were lower in the recurrence group than in the non-recurrence group (both P < 0.001). Predefined multivariable Cox regression analysis showed that LASr (per 1% increase: hazard ratio [HR] = 0.891, 95% confidence interval [CI]: 0.842–0.943, P < 0.001) and LAAV (per 1 cm/s increase: HR = 0.972, 95% CI: 0.956–0.989, P = 0.001) were independent predictors of AF recurrence after PFA; AF duration did not reach statistical significance after adjustment (HR = 1.005, 95% CI: 0.996–1.014, P = 0.268). The 18-month area under the curve (AUC) of the baseline clinical model constructed with persistent AF, AF duration, and left atrial volume index (LAVI) was 0.752 (95% CI: 0.687–0.817). After adding LASr and LAAV to the baseline model, the 18-month AUC increased to 0.891 (95% CI: 0.849–0.933), outperforming the baseline model (bootstrap P < 0.001) and improving reclassification ability (net reclassification index [NRI] = 0.438, 95% CI: 0.286–0.590; integrated discrimination improvement [IDI] = 0.158, 95% CI: 0.108–0.208; both P < 0.001). Based on LASr ≤ 22.3% and LAAV ≤ 45.2 cm/s, the 18-month recurrence-free rates in the low-, intermediate-, and high-risk groups were 88.7%, 70.5%, and 36.1%, respectively.

Conclusions

LASr and LAAV were independent predictors of AF recurrence after PFA, and their combination provided incremental predictive value beyond conventional clinical factors. Risk stratification based on LASr ≤ 22.3% and LAAV ≤ 45.2 cm/s may be used for preprocedural risk communication and assessment of postprocedural follow-up intensity; however, whether this approach can guide changes in ablation strategy and improve outcomes requires validation in prospective studies.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12872-026-06131-x.

Introduction

Atrial fibrillation (AF) is one of the most common sustained arrhythmias in clinical practice. Its prevalence increases substantially with age and is associated with increased risks of stroke, heart failure, and death [1]. Catheter ablation has become an important rhythm-control therapy for symptomatic AF. Pulsed field ablation (PFA), as a nonthermal ablation technique, has shown favorable efficacy and safety in prospective clinical studies [2]. Large-scale real-world data have demonstrated favorable overall periprocedural safety for PFA, with low risks of severe esophageal injury and persistent phrenic nerve palsy [3]. Compared with conventional thermal ablation, PFA has shown potential advantages in reducing the burden of postablation atrial arrhythmia, although the risk of recurrence has not been completely eliminated [4].

Postablation recurrence is related to multiple mechanisms, including pulmonary vein reconnection, progression of the atrial substrate, and non-pulmonary-vein triggers; therefore, preprocedural identification of recurrence risk remains important [5]. Peak left atrial reservoir strain (LASr) reflects left atrial reservoir function and is an important imaging marker for evaluating atrial mechanical remodeling [6]. Left atrial strain has potential clinical utility in multiple settings, including AF, heart failure, and valvular heart disease, but its measurement depends on image quality and software platform [7]. Previous studies have suggested that reduced left atrial reservoir function is associated with an increased risk of recurrence after AF ablation [8].

In contrast, reduced left atrial appendage flow velocity (LAAV) reflects impaired left atrial appendage emptying function and is associated with recurrence risk after ablation in patients with persistent AF [9]. Transesophageal echocardiography (TEE) combined with speckle-tracking parameters can provide complementary assessment of recurrence risk from the perspectives of left atrial appendage flow and mechanical function [10]. At present, studies on predictors of AF recurrence after PFA remain relatively limited, particularly comprehensive evaluations combining LASr and LAAV [11, 12]. This study aimed to investigate the predictive value of LASr combined with LAAV for AF recurrence after PFA, with the goal of providing evidence for preprocedural risk assessment and follow-up management.

Materials and methods

Study population

This study retrospectively analyzed clinical data from patients with AF who underwent PFA at our arrhythmia center between June 2021 and June 2024. The inclusion criteria were as follows: (1) age ≥ 18 years; (2) fulfillment of diagnostic criteria for AF, with AF diagnosed according to AF rhythm documented by 12-lead electrocardiogram (ECG) or ambulatory ECG and in accordance with current American and European AF management guidelines [13, 14]; (3) first PFA procedure; (4) complete preoperative TEE data with image quality suitable for analysis; (5) complete preoperative transthoracic echocardiography (TTE) with clear left atrial strain images; and (6) complete follow-up data. The exclusion criteria were as follows: (1) previous catheter ablation or surgical treatment; (2) valvular AF; (3) severe valvular disease, defined as moderate-to-severe valvular stenosis or regurgitation; (4) congenital heart disease; (5) left atrial thrombus; (6) malignancy or expected survival < 1 year; (7) poor echocardiographic image quality that precluded strain analysis; and (8) loss to follow-up. Definitions of the blanking period and recurrence endpoints after the blanking period were based on the international expert consensus statement on catheter and surgical ablation [15]. This study was approved by the Medical Ethics Committee of Wenzhou Hospital of Integrated Traditional Chinese and Western Medicine (approval No. 2026-L031) and complied with the ethical principles for medical research involving human participants in the Declaration of Helsinki [16]. As a retrospective study, the requirement for informed consent was waived.

Clinical data collection

Baseline clinical data were collected within 1 week before the procedure, including: (1) general information: age, sex, and body mass index (BMI); (2) AF type: paroxysmal AF or persistent AF; (3) AF duration; (4) comorbidities: hypertension, diabetes mellitus, coronary artery disease, heart failure, and history of stroke/transient ischemic attack (TIA); (5) CHA₂DS₂-VASc score, comprising congestive heart failure, hypertension, age ≥ 75 years, diabetes mellitus, stroke/transient ischemic attack/thromboembolism, vascular disease, age 65–74 years, and sex category; (6) preoperative medication use: anticoagulants and antiarrhythmic drugs; and (7) laboratory tests: complete blood count, liver and renal function, lipid profile, brain natriuretic peptide (BNP) or N-terminal pro-brain natriuretic peptide (NT-proBNP), and thyroid function. AF duration was defined as the time interval, in months, from the first definite diagnosis of AF by 12-lead ECG, ambulatory ECG, or verifiable medical record to the date of PFA. For patients first diagnosed at another hospital, the earliest traceable objective ECG record or medical-record diagnosis date was used; if only the year was available, the middle of that year was used as the estimated date.

Echocardiography

All patients underwent TTE within 1 week before the procedure and TEE within 24–72 h before the procedure. A Philips EPIQ 7 ultrasound system was used; TTE was performed with an X5-1 broadband matrix probe (frequency 1–5 MHz), and TEE was performed with an X8-2t matrix transesophageal probe. Patients were placed in the left lateral decubitus position, connected to ECG, and images were acquired at end-expiration during quiet breathing. Rhythm status at the time of each TTE and TEE examination was recorded and classified as sinus rhythm or AF rhythm. In patients in AF rhythm, left atrial strain and LAAV were measured in multiple representative cardiac cycles and averaged to reduce the influence of R-R interval variability.

Conventional echocardiographic parameters

Left ventricular ejection fraction (LVEF, measured using the biplane Simpson method), left atrial diameter (LAD), left atrial volume index (LAVI, calculated as maximum left atrial volume at end-ventricular systole measured using the biplane Simpson method divided by body surface area), interventricular septal thickness, left ventricular posterior wall thickness, mitral inflow spectrum (early diastolic peak velocity E wave, late diastolic peak velocity A wave, and E/A ratio), early diastolic mitral annular velocity (e′) at the septal and lateral annulus by tissue Doppler imaging, and E/e′ ratio were measured [17].

Left atrial strain analysis

Left atrial strain was evaluated using two-dimensional speckle-tracking echocardiography (2D-STE). High-frame-rate (70–90 frames/s) two-dimensional grayscale images from the apical four-chamber and two-chamber views were acquired over three consecutive cardiac cycles at end-expiration during quiet breathing. Clear visualization of the left atrial endocardial border was required, and the left atrial appendage and pulmonary veins were excluded. Offline analysis was performed using a QLAB workstation (Philips Healthcare). LASr measurement followed the recommendations for standardization of left atrial deformation imaging, with the onset of the QRS complex used as the zero reference point [18]. For patients in sinus rhythm, the onset of the QRS complex was used as the zero reference point; for patients with AF, cardiac cycles with R-R intervals close to the mean R-R interval (± 10%) and stable QRS-T morphology were selected for analysis. The left atrial endocardial border was manually traced, and the software automatically tracked atrial wall motion. The region of interest was adjusted to include the full thickness of the atrial wall. The software automatically divided the left atrium into six segments; segments with poor tracking quality were excluded, and a global left atrial longitudinal strain curve was obtained. Offline LASr analysis requires dedicated software, standardized region-of-interest adjustment, and quality control [19]. The measured parameters included LASr, left atrial conduit strain (LAScd), and left atrial contractile strain (LASct). Measurements from the four-chamber and two-chamber views were averaged as the final result.

LAAV measurement

During TEE, patients received local anesthesia or mild sedation. The left atrium and left atrial appendage were observed from multiple midesophageal angles to exclude thrombus or spontaneous echocardiographic contrast. In the 45°-90° view, a pulsed-wave Doppler sample volume (1–2 mm) was placed at the ostium of the left atrial appendage, approximately 1 cm from the orifice, to obtain left atrial appendage emptying and filling flow spectra. The primary measurements were peak left atrial appendage emptying velocity (referred to as LAAV in this article) and peak filling velocity. LAAV was measured as the peak left atrial appendage emptying velocity using TEE pulsed-wave Doppler and was averaged over representative cardiac cycles according to rhythm status [20]. In patients in sinus rhythm, emptying velocity was measured after the P wave over five consecutive cardiac cycles and averaged. In patients with AF, at least 10 cardiac cycles with R-R intervals within ± 10% of the mean R-R interval were selected, and LAAV was measured and averaged.

Pulsed field ablation procedure

Antiarrhythmic drugs were discontinued for at least five half-lives before the procedure. Regular anticoagulation was administered for at least 3 weeks preoperatively. In patients receiving warfarin, the international normalized ratio (INR) was maintained at 2.0–3.0; in patients receiving non-vitamin K antagonist oral anticoagulants (NOACs), the drug was discontinued 12–24 h before the procedure (after the last dose on the morning of the procedure). Patients fasted from food and water on the morning of the procedure. The procedure was performed under local anesthesia or general anesthesia. A coronary sinus electrode catheter was introduced through the right femoral vein. Transseptal puncture was performed through the right or bilateral femoral veins under fluoroscopic and ultrasound guidance. After successful puncture, intravenous heparin 100 U/kg was administered, and activated clotting time (ACT) was monitored intraoperatively and maintained at 300–350 s; additional heparin 1000–2000 U was administered hourly as needed to maintain the target ACT. A Farapulse PFA system was used, and ablation was performed with a pentaspline or basket ablation catheter under guidance of a three-dimensional electroanatomic mapping system.

Ablation strategy

All patients underwent pulmonary vein isolation (PVI), with circumferential ablation at the ostia of the four pulmonary veins. The number of pulse applications for each pulmonary vein was determined according to real-time monitoring of changes in pulmonary vein potentials and was usually 4–6 applications until pulmonary vein potentials disappeared. For patients with persistent AF, additional linear ablation (left atrial roof line and mitral isthmus line) or complex fractionated atrial electrogram (CFAE) ablation was performed at the operator’s discretion. The ablation strategy for each patient was recorded, including PVI alone, left atrial roof-line ablation, mitral isthmus-line ablation, CFAE ablation, and combined multi-strategy ablation. Procedure time, fluoroscopy time, and total number of pulse applications were also recorded.

Ablation endpoint

Acute complete electrical isolation of all four pulmonary veins was achieved, with disappearance of pulmonary vein potentials. Pulmonary vein isolation was reconfirmed 30 min after the final ablation. When necessary, 20–30 µg of isoproterenol was administered intravenously to observe whether pulmonary vein conduction recovered.

Postprocedural management

Anticoagulation was continued for at least 3 months after the procedure. Patients with a CHA₂DS₂-VASc score ≥ 1 in men or ≥ 2 in women received long-term anticoagulation. Antiarrhythmic drugs (AADs) were routinely used for 3 months after the procedure during the blanking-period management. After the 3-month blanking period, continuation of AADs was determined by the attending physician based on symptom burden, rhythm-monitoring results, drug tolerance, and recurrence risk. AADs were usually discontinued in patients without evidence of recurrence and with stable symptoms, whereas patients with persistent symptoms or a high risk of recurrence could continue treatment. Proton-pump inhibitors were used for 4–6 weeks to prevent gastrointestinal bleeding.

Follow-up and endpoint definitions

Outpatient or telephone follow-up was performed at 1, 3, 6, and 12 months after the procedure and every 6 months thereafter. At each follow-up visit, symptoms, 12-lead ECG results, and 24-h Holter findings were recorded. Patients with symptoms could undergo additional ECG or Holter examination at any time. The first 3 months after the procedure were defined as the blanking period. The follow-up cutoff date was uniformly set as December 30, 2024. The primary endpoint was AF recurrence, defined as documented AF, atrial flutter, or atrial tachycardia lasting ≥ 30 s after the blanking period, including symptomatic and asymptomatic episodes. Follow-up time was calculated from the date of PFA to the first recurrence, last valid follow-up, or the follow-up cutoff date. Patients without recurrence were censored at the last valid follow-up. Implantable ECG monitors or continuous wearable ECG monitoring were not used as routine follow-up tools in this study. To evaluate the balance of monitoring intensity, scheduled follow-up completion, number of scheduled ECG recordings, number of scheduled 24-h Holter recordings, and symptom-triggered additional ECG or Holter examinations were recorded for each patient. Secondary endpoints included: (1) periprocedural complications, including cardiac tamponade, stroke/TIA, vascular complications, esophageal injury, and phrenic nerve palsy; and (2) atrial arrhythmia episodes within 3 months after the procedure. Patients were divided into recurrence and non-recurrence groups according to the occurrence of AF recurrence during follow-up.

Statistical analysis

Statistical analyses were performed using SPSS 26.0 and R 4.2.1. Continuous variables with a normal distribution are expressed as mean ± standard deviation (x̄±s), and between-group comparisons were performed using the independent-samples t-test. Non-normally distributed continuous variables are expressed as median (interquartile range) [M(P₂₅, P₇₅)], and between-group comparisons were performed using the Mann-Whitney U test. Categorical variables are expressed as counts and percentages [n (%)], and between-group comparisons were performed using the χ² test or Fisher’s exact test. Univariate Cox analysis was reported as an exploratory analysis and was not used for automatic selection of variables into the main model. Based on clinical relevance and the study hypothesis, the main model was prespecified to include five variables: persistent AF, AF duration, LAVI, LASr, and LAAV. The proportional hazards assumption was assessed using Schoenfeld residuals. Spearman correlation analysis was used to evaluate correlations among LAVI, LASr, and LAAV. Variance inflation factors (VIFs) were further calculated in the main Cox model including persistent AF, AF duration, LAVI, LASr, and LAAV; VIF < 5 was considered to indicate no multicollinearity with a substantial influence. Model discrimination was evaluated using the 18-month time-dependent receiver operating characteristic (ROC) curve, with censoring handled by the inverse probability of censoring weighting (IPCW) method. When comparing the baseline clinical model with extended models, 1000 bootstrap resamples were used to calculate confidence intervals and P values for differences in area under the curve (AUC). The baseline clinical model included persistent AF, AF duration, and LAVI; extended models added LASr, LAAV, or both. Time-dependent net reclassification index (NRI) and integrated discrimination improvement (IDI) were used to evaluate incremental predictive value. Survival analysis was performed using the Kaplan-Meier method, and recurrence-free survival rates were compared using the log-rank test. To assess model robustness, 1000 bootstrap resamples were used to calculate the optimism-corrected concordance index (C-index), calibration slope, and 18-month Brier score, and ridge-penalized Cox regression was performed as a sensitivity analysis. All tests were two-sided, and P < 0.05 was considered statistically significant.

Results

General characteristics

A total of 246 patients with AF who underwent PFA were included in this study, including 156 men (63.4%) and 90 women (36.6%), with a mean age of 61.8 ± 10.3 years. There were 182 patients (74.0%) with paroxysmal AF and 64 patients (26.0%) with persistent AF. As of December 30, 2024, the median follow-up duration was 17.3 (10.8, 25.6) months, the mean follow-up duration was 17.8 ± 8.6 months, and the follow-up range was 6.0-42.8 months. During follow-up, 68 patients (27.6%) experienced recurrence of atrial arrhythmia after the blanking period. The median time to recurrence in the recurrence group was 9.8 (5.9, 14.3) months, and the mean time to recurrence was 10.5 ± 5.3 months.

The recurrence group had higher proportions of persistent AF and heart failure, longer AF duration, higher CHA₂DS₂-VASc scores, and higher NT-proBNP levels than the non-recurrence group (all P < 0.05). Age, sex, BMI, hypertension, diabetes mellitus, coronary artery disease, history of stroke/TIA, preoperative anticoagulant use, preoperative antiarrhythmic drug use, hemoglobin, white blood cell count, platelet count, serum creatinine, estimated glomerular filtration rate (eGFR), alanine aminotransferase (ALT), aspartate aminotransferase (AST), total cholesterol, triglycerides, low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), thyroid-stimulating hormone (TSH), free triiodothyronine (FT3), and free thyroxine (FT4) did not differ significantly between groups (all P > 0.05) (Table 1).

Table 1.

Comparison of baseline clinical characteristics between the recurrence and non-recurrence groups

Variable	Recurrence group (n = 68)	Non-recurrence group (n = 178)	Statistic	P value
Age (years)	63.2 ± 9.8	61.2 ± 10.5	t = 1.358	0.176
Male [n (%)]	45 (66.2)	111 (62.4)	χ²=0.308	0.579
BMI (kg/m²)	25.8 ± 3.2	25.3 ± 3.5	t = 1.021	0.308
Persistent AF [n (%)]	27 (39.7)	37 (20.8)	χ²=9.124	0.003
AF duration [months, M(P₂₅, P₇₅)]	36.0 (18.0, 60.0)	24.0 (12.0, 48.0)	Z=-2.512	0.012
Hypertension [n (%)]	42 (61.8)	98 (55.1)	χ²=0.903	0.342
Diabetes mellitus [n (%)]	16 (23.5)	34 (19.1)	χ²=0.609	0.435
Coronary artery disease [n (%)]	12 (17.6)	26 (14.6)	χ²=0.365	0.546
Heart failure [n (%)]	11 (16.2)	12 (6.7)	χ²=4.838	0.028
History of stroke/TIA [n (%)]	8 (11.8)	14 (7.9)	χ²=0.948	0.330
CHA₂DS₂-VASc score [M(P₂₅, P₇₅)]	3.0 (2.0, 4.0)	2.0 (1.0, 3.0)	Z=-2.748	0.006
NT-proBNP [pg/mL, M(P₂₅, P₇₅)]	186.5 (112.3, 298.7)	128.4 (76.5, 201.2)	Z=-2.891	0.004
Preoperative medication
Anticoagulants [n (%)]	65 (95.6)	168 (94.4)	χ²=0.122	0.726
Antiarrhythmic drugs [n (%)]	48 (70.6)	118 (66.3)	χ²=0.399	0.527
Laboratory tests
Hemoglobin (g/L)	138.5 ± 15.6	140.2 ± 14.8	t=-0.798	0.426
White blood cell count (×10⁹/L)	6.8 ± 1.9	6.6 ± 1.8	t = 0.762	0.447
Platelet count (×10⁹/L)	218.4 ± 56.3	224.6 ± 58.7	t=-0.758	0.449
Serum creatinine (µmol/L)	78.6 ± 18.4	76.2 ± 17.2	t = 0.964	0.336
eGFR (mL/min/1.73 m²)	88.5 ± 16.8	90.8 ± 15.4	t=-1.023	0.307
ALT (U/L)	28.6 ± 12.5	26.8 ± 11.3	t = 1.082	0.280
AST (U/L)	26.4 ± 9.8	25.2 ± 8.6	t = 0.938	0.349
Total cholesterol (mmol/L)	4.52 ± 0.86	4.48 ± 0.92	t = 0.318	0.751
Triglycerides [mmol/L, M(P₂₅, P₇₅)]	1.45 (1.08, 2.12)	1.38 (1.02, 1.98)	Z=-0.865	0.387
LDL-C (mmol/L)	2.68 ± 0.74	2.62 ± 0.78	t = 0.561	0.575
HDL-C (mmol/L)	1.18 ± 0.32	1.22 ± 0.35	t=-0.842	0.401
TSH [mIU/L, M(P₂₅, P₇₅)]	2.18 (1.42, 3.26)	2.05 (1.35, 3.08)	Z=-0.742	0.458
FT3 (pmol/L)	4.86 ± 0.68	4.92 ± 0.72	t=-0.608	0.544
FT4 (pmol/L)	16.8 ± 2.4	17.1 ± 2.6	t=-0.848	0.397

Continuous variables are expressed as mean ± SD or median (IQR); categorical variables are expressed as n (%). t, Z, and χ² indicate independent-samples t-test, Mann-Whitney U test, and chi-square test statistics, respectively; P < 0.05 was considered statistically significant

BMI body mass index, AF atrial fibrillation, TIA transient ischemic attack, CHA₂DS₂-VASc congestive heart failure, hypertension, age ≥ 75 years, diabetes mellitus, stroke/transient ischemic attack/thromboembolism, vascular disease, age 65–74 years, sex category, AAD antiarrhythmic drug, eGFR estimated glomerular filtration rate, ALT alanine aminotransferase, AST aspartate aminotransferase, LDL-C low-density lipoprotein cholesterol, HDL-C high-density lipoprotein cholesterol, TSH thyroid-stimulating hormone, FT3 free triiodothyronine, FT4 free thyroxine, BNP brain natriuretic peptide, NT-proBNP N-terminal pro-brain natriuretic peptide, SD standard deviation, IQR interquartile range

Comparison of echocardiographic parameters

LAD, LAVI, and E/e′ were higher in the recurrence group than in the non-recurrence group, whereas E/A, LASr, LAScd, LASct, and LAAV were lower in the recurrence group (all P < 0.05). LVEF, interventricular septal thickness, left ventricular posterior wall thickness, E wave, A wave, and e′ did not differ significantly between groups (all P > 0.05) (Table 2). The proportions of patients in AF rhythm during TTE and TEE were higher in the recurrence group than in the non-recurrence group (both P < 0.05). After further adjustment for rhythm status during TTE and TEE, the direction and statistical significance of the associations of LASr and LAAV with recurrence remained consistent (Supplementary Tables 7 and 8).

Table 2.

Comparison of echocardiographic parameters between the recurrence and non-recurrence groups

Variable	Recurrence group (n = 68)	Non-recurrence group (n = 178)	Statistic	P value
LVEF (%)	61.8 ± 6.2	62.5 ± 5.8	t=-0.835	0.405
LAD (mm)	43.2 ± 4.8	39.6 ± 4.2	t = 5.641	< 0.001
LAVI [mL/m², M(P₂₅, P₇₅)]	38.5 (32.4, 46.2)	32.1 (27.3, 37.8)	Z=-4.523	< 0.001
Interventricular septal thickness (mm)	9.8 ± 1.2	9.6 ± 1.1	t = 1.198	0.232
Left ventricular posterior wall thickness (mm)	9.5 ± 1.1	9.3 ± 1.0	t = 1.324	0.187
E wave (cm/s)	78.5 ± 16.2	76.2 ± 15.4	t = 1.015	0.311
A wave (cm/s)	68.2 ± 18.4	70.2 ± 18.8	t=-0.762	0.447
E/A	0.98 ± 0.28	1.12 ± 0.30	t=-3.362	0.001
e′ (cm/s)	6.8 ± 1.5	7.2 ± 1.6	t=-1.782	0.076
E/e′	11.8 ± 2.6	10.2 ± 2.3	t = 4.586	< 0.001
LASr (%)	18.2 ± 4.3	26.5 ± 5.1	t=-11.976	< 0.001
LAScd (%)	9.8 ± 3.1	14.2 ± 3.6	t=-8.978	< 0.001
LASct (%)	8.4 ± 2.7	12.3 ± 3.2	t=-8.962	< 0.001
LAAV (cm/s)	38.5 ± 12.4	52.8 ± 14.6	t=-7.176	< 0.001

Continuous variables are expressed as mean ± SD or median (IQR); categorical variables are expressed as n (%). t, Z, and χ² indicate independent-samples t-test, Mann-Whitney U test, and chi-square test statistics, respectively

LVEF left ventricular ejection fraction, LAD left atrial diameter, LAVI left atrial volume index, LASr left atrial reservoir strain, LAScd left atrial conduit strain, LASct left atrial contractile strain, LAAV left atrial appendage flow velocity, TTE transthoracic echocardiography, TEE transesophageal echocardiography, SD standard deviation, IQR interquartile range

Procedural data

PFA was successfully completed in all patients, and the acute PVI success rate was 100%. The proportion of PVI alone was lower in the recurrence group than in the non-recurrence group, whereas the proportion of any additional ablation, procedure time, and total number of pulse applications were higher in the recurrence group (all P < 0.05). Single left atrial roof-line ablation, single mitral isthmus-line ablation, single CFAE ablation, combined two additional strategies, combined three additional strategies, fluoroscopy time, periprocedural complications, blanking-period atrial arrhythmia, and continued AAD use after 3 months did not differ significantly between groups (all P > 0.05) (Table 3).

Table 3.

Procedural data and stratified comparison by recurrence status

Variable	Overall (n = 246)	Recurrence group (n = 68)	Non-recurrence group (n = 178)	P value
Acute PVI success [n (%)]	246 (100)	68 (100)	178 (100)	Not applicable
PVI alone [n (%)]	195 (79.3)	45 (66.2)	150 (84.3)	0.002
Any additional ablation [n (%)]	51 (20.7)	23 (33.8)	28 (15.7)	0.002
Single left atrial roof-line ablation [n (%)]	20 (8.1)	9 (13.2)	11 (6.2)	0.114
Single mitral isthmus-line ablation [n (%)]	9 (3.7)	4 (5.9)	5 (2.8)	0.266
Single CFAE ablation [n (%)]	4 (1.6)	2 (2.9)	2 (1.1)	0.306
Two additional strategies combined [n (%)]	16 (6.5)	7 (10.3)	9 (5.1)	0.153
Three additional strategies combined [n (%)]	2 (0.8)	1 (1.5)	1 (0.6)	0.477
Procedure time (min)	98.5 ± 22.3	105.4 ± 23.6	95.9 ± 21.2	0.004
Fluoroscopy time (min)	18.6 ± 5.4	19.7 ± 5.6	18.2 ± 5.2	0.058
Total number of pulse applications	42.8 ± 8.6	45.9 ± 9.2	41.6 ± 8.1	0.001
Periprocedural complications [n (%)]	7 (2.8)	3 (4.4)	4 (2.2)	0.399
Blanking-period atrial arrhythmia [n (%)]	70 (28.5)	22 (32.4)	48 (27.0)	0.404
Continued AAD use after 3 months [n (%)]	57 (23.2)	19 (27.9)	38 (21.3)	0.273

Categorical variables are expressed as n (%), and continuous variables as mean ± standard deviation or median (interquartile range). P values were derived from the χ² test, Fisher’s exact test, independent-samples t-test, or Mann-Whitney U test, as appropriate

PFA pulsed field ablation, PVI pulmonary vein isolation, CFAE complex fractionated atrial electrogram, ACT activated clotting time, AAD antiarrhythmic drug, ECG electrocardiogram, Holter 24-h ambulatory electrocardiography

The scheduled follow-up completion rate was 95.9%. There were no significant differences between the recurrence and non-recurrence groups in scheduled follow-up completion rate, number of scheduled ECG recordings, or number of scheduled 24-h Holter recordings. The proportion of symptom-triggered additional ECG or Holter examinations was higher in the recurrence group (Supplementary Table 3). A total of 57 patients continued AADs after 3 months. The proportion of continued AAD use was higher in the recurrence group than in the non-recurrence group, but the difference was not statistically significant (P > 0.05) (Table 3 and Supplementary Table 9). A total of 70 patients experienced atrial arrhythmia during the blanking period. The trend toward a higher incidence of blanking-period atrial arrhythmia in the recurrence group was not prominent, and the between-group difference was not statistically significant (P > 0.05) (Table 3).

Univariate Cox regression analysis of AF recurrence

Univariate Cox analysis showed that persistent AF, AF duration, heart failure, CHA₂DS₂-VASc score, NT-proBNP, LAD, LAVI, E/A, E/e′, LASr, LAScd, LASct, and LAAV were associated with recurrence after the blanking period (all P < 0.05). Age, sex, BMI, hypertension, diabetes mellitus, coronary artery disease, history of stroke/TIA, preoperative medication use, LVEF, and conventional laboratory indicators were not significantly associated with recurrence (all P > 0.05) (Table 4).

Table 4.

Univariate Cox regression analysis of AF recurrence

Variable	HR	95% CI	P value
Age	1.018	0.997–1.040	0.092
Male	1.176	0.715–1.934	0.523
BMI	1.042	0.967–1.122	0.278
Persistent AF	2.184	1.326–3.598	0.002
AF duration	1.012	1.004–1.021	0.005
Hypertension	1.316	0.810–2.138	0.268
Diabetes mellitus	1.285	0.738–2.237	0.376
Coronary artery disease	1.248	0.676–2.305	0.478
Heart failure	2.156	1.145–4.060	0.017
History of stroke/TIA	1.524	0.735–3.160	0.258
CHA₂DS₂-VASc score	1.285	1.092–1.512	0.003
NT-proBNP (per 100 pg/mL increase)	1.286	1.092–1.515	0.003
Anticoagulants	0.876	0.324–2.368	0.792
Antiarrhythmic drugs	1.214	0.726–2.030	0.459
LVEF	0.985	0.950–1.022	0.422
LAD	1.158	1.092–1.228	< 0.001
LAVI	1.064	1.037–1.092	< 0.001
E/A	0.412	0.235–0.723	0.002
E/e′	1.176	1.088–1.271	< 0.001
LASr	0.848	0.807–0.890	< 0.001
LAScd	0.876	0.825–0.931	< 0.001
LASct	0.857	0.799–0.920	< 0.001
LAAV	0.961	0.947–0.975	< 0.001

HRs were calculated using univariate Cox proportional hazards models

HR hazard ratio, CI confidence interval, BMI body mass index, AF atrial fibrillation, TIA transient ischemic attack, CHA₂DS₂-VASc congestive heart failure, hypertension, age ≥ 75 years, diabetes mellitus, stroke/transient ischemic attack/thromboembolism, vascular disease, age 65–74 years, sex category, NT-proBNP N-terminal pro-brain natriuretic peptide, AAD antiarrhythmic drug, LVEF left ventricular ejection fraction, LAD left atrial diameter, LAVI left atrial volume index, LASr left atrial reservoir strain, LAScd left atrial conduit strain, LASct left atrial contractile strain, LAAV left atrial appendage flow velocity

Multivariable Cox regression analysis of AF recurrence

In the predefined multivariable Cox model, persistent AF, LAVI, LASr, and LAAV remained independently associated with AF recurrence after PFA. AF duration was associated with recurrence in univariate analysis, but no longer reached statistical significance after simultaneous adjustment for persistent AF, LAVI, LASr, and LAAV. The Schoenfeld residual test showed that the main model satisfied the proportional hazards assumption (overall P = 0.462) (Table 5).

Table 5.

Predefined multivariable Cox regression analysis of AF recurrence

Variable	HR	95% CI	P value	VIF
Persistent AF	1.742	1.026–2.957	0.040	1.34
AF duration (per 1-month increase)	1.005	0.996–1.014	0.268	1.29
LAVI (per 1 mL/m² increase)	1.029	1.003–1.056	0.030	1.57
LASr (per 1% increase)	0.891	0.842–0.943	< 0.001	1.76
LAAV (per 1 cm/s increase)	0.972	0.956–0.989	0.001	1.42

HRs were calculated using the predefined multivariable Cox proportional hazards model; VIF < 5 indicated no multicollinearity with a substantial influence

HR hazard ratio, CI confidence interval, VIF variance inflation factor, AF atrial fibrillation, LAVI left atrial volume index, LASr left atrial reservoir strain, LAAV left atrial appendage flow velocity

Moderate correlations were observed among LAVI, LASr, and LAAV, but the magnitude did not suggest severe collinearity. LAVI was negatively correlated with LASr and LAAV, whereas LASr was positively correlated with LAAV (all P < 0.001). All variables in the main Cox model had VIF values < 2.0, suggesting that model estimates were not affected by severe multicollinearity (Supplementary Tables 1 and 2).

Incremental predictive value of LASr and LAAV over the baseline clinical model

The 18-month AUC of the baseline clinical model was 0.752. After adding LASr or LAAV to the baseline model, the AUC increased. After adding both LASr and LAAV, the AUC further increased to 0.891 and was superior to that of the baseline model (bootstrap P < 0.001). The combined extended model also improved reclassification ability (NRI = 0.438; IDI = 0.158; both P < 0.001) (Table 6 and Fig. 1).

Table 6.

Incremental predictive value of LASr and LAAV over the baseline clinical model (18-month time-dependent ROC)

Model	Variables included	18-month AUC (95% CI)	ΔAUC vs. baseline	Bootstrap P value	Time-dependent NRI; IDI	Sensitivity (%)	Specificity (%)
LASr-only model	LASr	0.818 (0.760–0.876)	Not applicable	Not applicable	Not applicable	79.4	76.4
LAAV-only model	LAAV	0.792 (0.731–0.853)	Not applicable	Not applicable	Not applicable	75.0	74.2
Baseline clinical model	Persistent AF + AF duration + LAVI	0.752 (0.687–0.817)	Reference	Reference	Reference	70.6	71.3
Baseline + LASr	Baseline clinical model + LASr	0.836 (0.783–0.889)	0.084	< 0.001	NRI = 0.302 (0.174–0.430); IDI = 0.086 (0.049–0.123)	80.9	77.0
Baseline + LAAV	Baseline clinical model + LAAV	0.819 (0.762–0.876)	0.067	0.003	NRI = 0.238 (0.112–0.364); IDI = 0.063 (0.031–0.095)	76.5	76.4
Baseline + LASr + LAAV	Baseline clinical model + LASr + LAAV	0.891 (0.849–0.933)	0.139	< 0.001	NRI = 0.438 (0.286–0.590); IDI = 0.158 (0.108–0.208)	85.3	82.0

AUCs are 18-month IPCW time-dependent ROC AUCs; bootstrap P values were obtained from 1000 resampling comparisons

AUC area under the curve, CI confidence interval, ΔAUC difference in area under the curve versus the baseline clinical model, IPCW inverse probability of censoring weighting, ROC receiver operating characteristic, NRI net reclassification index, IDI integrated discrimination improvement, AF atrial fibrillation, LAVI left atrial volume index, LASr left atrial reservoir strain, LAAV left atrial appendage flow velocity

Fig. 1.

Inverse probability of censoring weighting (IPCW) time-dependent receiver operating characteristic (ROC) curves for predicting 18-month AF recurrence using the baseline clinical model and extended models. The baseline model included persistent AF, AF duration, and LAVI; the extended models added LASr, LAAV, or both. AF, atrial fibrillation; LAVI, left atrial volume index; LASr, left atrial reservoir strain; LAAV, left atrial appendage flow velocity; AUC, area under the curve

Subgroup analysis and risk stratification

Subgroup analysis according to AF type showed that in patients with paroxysmal AF, both LASr and LAAV were lower in the recurrence group than in the non-recurrence group (both P < 0.001). In patients with persistent AF, both LASr and LAAV were also lower in the recurrence group than in the non-recurrence group (both P < 0.05). In exploratory Cox analysis among patients with persistent AF, LASr and LAAV remained associated with recurrence risk (Supplementary Table 6).

Based on LASr ≤ 22.3% and LAAV ≤ 45.2 cm/s, patients were divided into three groups: LASr > 22.3% and LAAV > 45.2 cm/s (low-risk group, n = 112); only LASr ≤ 22.3% or only LAAV ≤ 45.2 cm/s (intermediate-risk group, n = 89); and LASr ≤ 22.3% and LAAV ≤ 45.2 cm/s (high-risk group, n = 45). Recurrence-free rates decreased progressively across the low-, intermediate-, and high-risk groups, and the between-group difference was statistically significant (log-rank P < 0.001) (Table 7 and Fig. 2).

Table 7.

AF recurrence according to risk-stratification group (Kaplan-Meier estimates)

Group	No. of patients	Total recurrences (%)	12-month recurrence-free rate (%)	No. at risk at 12 months	18-month recurrence-free rate (%)	No. at risk at 18 months
Low-risk group: LASr > 22.3% and LAAV > 45.2 cm/s	112	13 (11.6)	92.1	98	88.7	76
Intermediate-risk group: only LASr ≤ 22.3% or only LAAV ≤ 45.2 cm/s	89	26 (29.2)	78.4	63	70.5	47
High-risk group: LASr ≤ 22.3% and LAAV ≤ 45.2 cm/s	45	29 (64.4)	55.9	22	36.1	10

The 12- and 18-month recurrence-free rates are Kaplan-Meier estimates. The number at risk refers to patients who remained under follow-up and had not experienced the endpoint before the corresponding time point. Between-group comparisons were performed using the log-rank test

LASr left atrial reservoir strain, LAAV left atrial appendage flow velocity

Fig. 2.

Kaplan-Meier recurrence-free survival curves stratified by LASr and LAAV cutoffs. Recurrence-free survival decreased sequentially across the low-, intermediate-, and high-risk groups. LASr, left atrial reservoir strain; LAAV, left atrial appendage flow velocity

Development and validation of the prediction model

A prediction model was developed based on the predefined multivariable Cox model. The apparent concordance index (C-index) was 0.872, and the optimism-corrected C-index after 1000 bootstrap resamples was 0.854. The calibration slope was 0.928, and the 18-month Brier score was 0.148. Decision curve analysis showed that when the threshold probability ranged from 12% to 68%, the combined model provided higher net benefit than strategies of classifying all patients as high risk or all patients as low risk. Ridge-penalized Cox sensitivity analysis showed that the direction and statistical significance of the effects of LASr and LAAV were consistent with the main model (Table 8 and Fig. 3, and Supplementary Table 4). After adjustment for procedural heterogeneity, procedure time, and total number of pulse applications, LASr and LAAV remained independently associated with recurrence risk (Supplementary Table 5).

Table 8.

Validation metrics of the prediction model

Metric	Value
Apparent C-index	0.872 (95% CI: 0.827–0.917)
Bootstrap optimism-corrected C-index	0.854
Calibration slope	0.928
18-month Brier score	0.148
Threshold range with net benefit in decision curve analysis	12%-68%

The C-index was used to evaluate model discrimination. The apparent C-index refers to the uncorrected model discrimination, whereas the bootstrap optimism-corrected C-index refers to internally validated discrimination. The Brier score was used to evaluate prediction error; lower values indicate smaller prediction error. The calibration slope was used to evaluate agreement between predicted probabilities and observed outcomes. Net benefit was derived from DCA. Bootstrap internal validation was performed with 1000 resamples

CI confidence interval, C-index concordance index, DCA decision curve analysis

Fig. 3.

Decision curve analysis (DCA) of the combined model. The combined model provided higher net benefit than strategies of classifying all patients as high risk or all patients as low risk across threshold probabilities of 12%-68%. The shaded area represents the threshold-probability range with net-benefit advantage. DCA, decision curve analysis

Discussion

This study systematically evaluated the predictive value of LASr combined with LAAV for AF recurrence after PFA. Among 246 patients with AF who underwent PFA, the recurrence rate as of December 30, 2024, was 27.6%. After adjustment for persistent AF, AF duration, and LAVI, LASr and LAAV remained independent predictors of recurrence, and their combined use further improved 18-month recurrence-risk discrimination and reclassification beyond the baseline clinical model. PFA studies with continuous ECG monitoring have suggested that more intensive rhythm monitoring can improve recurrence detection, which is consistent with the cautious interpretation of follow-up sensitivity adopted in this study [21].

Blanking-period arrhythmia did not reach statistical significance in this study, but this does not mean that early arrhythmia lacks predictive significance. Possible explanations include limited statistical power due to the number of events, limited detection of brief or asymptomatic episodes by fixed-time-point ECG/Holter monitoring, and routine AAD use during the blanking period, which may reduce early arrhythmia burden. A related analysis from the admIRE study showed that early recurrence after PFA was associated with an increased risk of late recurrence; therefore, the present findings should be interpreted as no independent association being observed under the current monitoring strategy and sample size rather than as a refutation of existing evidence [22].

Procedural heterogeneity is an important factor that must be considered when interpreting recurrence risk after PFA. A direct epicardial validation study supports the ability of PFA to produce an electroporation effect on the posterior wall in persistent AF, providing mechanistic evidence for substrate modification [23]. A comparative study of pentaspline PFA versus high-power short-duration or very-high-power short-duration radiofrequency ablation suggests that differences in energy platforms and ablation extent may affect outcome interpretation [24]. In this study, the recurrence group had higher proportions of additional ablation, longer procedure time, and more pulse applications, suggesting that these procedural indicators may reflect a more complex atrial substrate. After adjustment for additional ablation, procedure time, and total number of pulse applications, LASr and LAAV remained associated with recurrence, supporting that their predictive value was not fully explained by procedural differences.

The mechanisms of recurrence after PFA may still involve multiple factors, including pulmonary vein reconnection, atrial tachycardia, and non-pulmonary-vein substrate. Studies of repeat procedures have shown that pulmonary vein reconnection and atrial tachycardia mechanisms can still be observed in patients with recurrence after PFA [25]. Studies of multielectrode PFA in persistent and long-standing persistent AF have emphasized the importance of individualized left atrial target ablation strategies [26]. Preliminary experience with variable-loop catheters also suggests that differences in PFA platforms and catheter design may affect procedural workflow and lesion formation [27]. Therefore, incorporating procedural strategy into sensitivity analyses in this study helps reduce the influence of procedural heterogeneity on the main conclusions.

LASr reflects left atrial reservoir function and complements conventional structural indices from the perspective of mechanical function. Three-dimensional speckle-tracking studies support the value of left atrial strain in predicting recurrence after AF ablation [28]. Studies in nonvalvular AF also suggest that speckle-tracking parameters may provide complementary information for recurrence-risk identification [29]. Three-dimensional echocardiographic studies have shown that comprehensive assessment of left atrial morphology, function, and strain helps explain differences in recurrence after pulmonary vein isolation [30]. In the present study, LAVI, LASr, and LAAV showed only moderate correlations, and all VIF values in the main model were < 2.0, indicating that LASr did not simply duplicate information on left atrial enlargement but provided reservoir-function information beyond structural remodeling.

Reduced LAAV indicates impaired left atrial appendage emptying function and decreased atrial mechanical function, and may represent a more severe atrial substrate phenotype. Lower preprocedural left atrial appendage flow velocity has been associated with adverse long-term events after AF ablation, suggesting that LAAV reflects not only thromboembolic risk but also impaired atrial mechanical function [31]. An early classic study showed that reduced LAAV in patients with persistent AF predicted recurrence after ablation, supporting the directionally consistent findings for LAAV in the present study [32]. A study of a left atrial appendage emptying velocity prediction model in patients with nonvalvular AF suggested that left atrial appendage emptying velocity is jointly influenced by atrial structure and clinical factors [33].

Left atrial appendage structure and function may jointly contribute to the formation of the atrial substrate phenotype. A systematic review of the association between left atrial appendage volume and recurrence after AF ablation suggested that left atrial appendage structure and function may jointly contribute to recurrence risk [34]. Multifactorial studies have shown that LAD, NT-proBNP, LAAV, and early recurrence can all contribute to recurrence-risk stratification [35]. In this study, LAAV remained associated with recurrence after adjustment for LASr, LAVI, AF type, and AF duration, suggesting that it can provide independent hemodynamic information for a comprehensive prediction model.

From the perspective of clinical implementation, LASr can be obtained from high-quality TTE images through offline 2D-STE analysis; however, its accuracy depends on image quality, analysis software, operator training, and quality control. Reviews of left atrial strain applications have noted that LASr has the advantage of reflecting atrial functional reserve, but its clinical implementation still requires standardized measurement procedures [36]. LAAV usually requires TEE measurement. Although many patients scheduled for AF ablation undergo preoperative TEE to exclude left atrial or left atrial appendage thrombus, TEE remains a semi-invasive examination and is limited by patient tolerance, sedation risk, and institutional availability. Differences in postprocedural premature atrial contraction burden among different ablation energy modalities suggest that the intensity of rhythm monitoring can influence estimates of recurrence burden [37]. Therefore, this model is most suitable for centers with standardized left atrial strain analysis, TEE workflows, and structured follow-up monitoring.

The clinical significance of this study should be limited to preprocedural risk assessment and follow-up management. The model combining LASr and LAAV can help identify patients at higher risk of recurrence after PFA, thereby informing preprocedural risk communication, planning of recurrence-monitoring frequency, and follow-up for anticoagulation and rhythm management. However, this study did not evaluate whether changing the ablation extent, energy strategy, or selecting alternative therapy based on this model can improve clinical outcomes. Therefore, at present, the results of LASr and LAAV alone should not be used to decide on expanding the ablation extent or altering the treatment pathway. A recent review emphasized that recurrence after AF ablation results from multiple mechanisms, and no single indicator can fully explain recurrence risk [38].

This study has several limitations. First, it was a single-center retrospective study with a relatively limited sample size; multicenter prospective studies are needed for further validation. Second, implantable loop recorders or continuous long-term ECG monitoring were not used during follow-up, and asymptomatic, brief, or low-burden atrial arrhythmias may have been missed. Although scheduled follow-up and scheduled Holter monitoring were similar between groups, the recurrence group had a higher proportion of symptom-triggered additional examinations, which may have introduced some degree of differential detection bias. Third, the clinical accessibility of LASr and LAAV should be interpreted cautiously. LASr is not automatically available in all routine echocardiographic examinations and requires dedicated speckle-tracking software and standardized training; LAAV requires TEE and cannot be regarded as a fully noninvasive parameter. Fourth, although this study used a prespecified Cox model, bootstrap internal validation, and multiple sensitivity analyses, these approaches cannot replace external validation. Fifth, this study did not include higher-level imaging parameters such as cardiac magnetic resonance assessment of left atrial fibrosis.

Conclusions

LASr and LAAV were independent predictors of AF recurrence after PFA, and their combination provided incremental predictive value beyond conventional clinical factors. Risk stratification based on LASr ≤ 22.3% and LAAV ≤ 45.2 cm/s may be used for preprocedural risk communication and assessment of postprocedural follow-up intensity; however, whether this approach can guide changes in ablation strategy and improve outcomes requires validation in prospective studies.

Authors’ contributions

SW and JM contributed equally to this work as co-first authors. SW and JM conceived and designed the study, participated in data collection, and drafted the manuscript. FT contributed to patient data collection in the physical examination center and assisted in data management. LW performed echocardiographic examinations and left atrial strain analysis. WC conducted transesophageal echocardiography and left atrial appendage flow velocity measurements. LQ assisted in data interpretation and statistical analysis. MZ supervised the entire study, provided critical intellectual input, and revised the manuscript. All authors read and approved the final manuscript.

Funding

This study received no specific funding support.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

All procedures performed in this study involving human participants were in accordance with the ethical standards of the institutional research committee and with the Declaration of Helsinki. This study was reviewed and approved by the Medical Ethics Committee of Wenzhou Hospital of Integrated Traditional Chinese and Western Medicine (approval No. 2026-L031). As this was a retrospective study, the requirement for individual informed consent was waived by the ethics committee.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.