Low-density lipoprotein cholesterol is inversely associated with left atrial low-voltage areas in patients with persistent atrial fibrillation: a retrospective cohort study Authors: Jiahe Hu, Beining E, Jiawei Lu, Wenxing Guo, Binli Xu, Xinpeng You, Ling Yang, Qi Jiang Published: 2026 ABSTRACT AbstractBackgroundLeft atrial low-voltage areas (LVAs) reflect advanced atrial remodeling and are associated with postablation recurrence in patients with persistent atrial fibrillation (PerAF). However, the relationship between routinely measured metabolic parameters and LVAs remains insufficiently characterized. This study aimed to identify clinical factors associated with LVAs in patients with PerAF, with particular emphasis on metabolic parameters.MethodsThis retrospective single-center cohort included consecutive PerAF patients who underwent initial catheter ablation between January 2021 and August 2024. LVA status was analyzed as a binary outcome based on the results of the intraprocedural substrate assessment. Logistic regression was used to assess associations with LVAs, while mediation, restricted cubic spline, and quartile-based analyses were conducted as secondary analyses; the quartile-based analyses were descriptive. Postablation recurrence was assessed by Kaplan–Meier and Cox analyses.ResultsAmong 404 patients, 232 (57.4%) had LVAs. Age, female sex, fibrillatory wave amplitude, left atrial diameter (LAD), and LDL-C levels were associated with the presence of LVAs. Higher LDL-C levels were associated with lower odds of LVAs (odds ratio [OR] 0.72 per mmol/L, 95% confidence interval [CI] 0.53–0.97). LAD may partly account for this association (15%–18%). Quartile-based analyses revealed heterogeneity in LVA prevalence and recurrence across LDL-C categories, but these results were descriptive.ConclusionIn patients with PerAF, higher LDL-C levels were independently associated with a lower likelihood of LVAs. Lower LDL-C levels may help identify patients with a higher likelihood of advanced atrial remodeling before ablation.Graphical Abstract Abbreviations: AF, atrial fibrillation; LAD, left atrial diameter; LDL-C, low-density lipoprotein cholesterol; LVA, low-voltage area; OR, odds ratio; Q1–Q4, quartiles 1–4 Supplementary InformationThe online version contains supplementary material available at 10.1186/s12944-026-03020-2.Keywords: Atrial fibrillation, left atrial low-voltage areas, Low-density lipoprotein cholesterol, Catheter ablation What’s new? Higher LDL-C levels were associated with a lower likelihood of left atrial low-voltage areas in patients with persistent atrial fibrillation, suggesting a possible association with atrial substrate remodeling. Higher LDL-C levels were associated with a lower likelihood of left atrial low-voltage areas in patients with persistent atrial fibrillation, suggesting a possible association with atrial substrate remodeling. Quartile-based analyses revealed descriptive heterogeneity in LVA prevalence and recurrence across LDL-C categories, but a formal range-dependent effect was not established. Left atrial diameter may partly account for the observed association between LDL-C levels and LVA presence. Quartile-based analyses revealed descriptive heterogeneity in LVA prevalence and recurrence across LDL-C categories, but a formal range-dependent effect was not established. Left atrial diameter may partly account for the observed association between LDL-C levels and LVA presence. Introduction Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia encountered in clinical practice and is associated with adverse outcomes, including ischemic stroke, heart failure, and increased overall mortality [1]. Although catheter ablation techniques have improved, long-term recurrence remains common in patients with persistent AF (PerAF) [2, 3]. The maintenance and recurrence of PerAF are closely linked to left atrial remodeling, of which fibrosis is a key feature [4, 5]. Compared with patients with paroxysmal AF, patients with PerAF usually have more advanced left atrial fibrosis and electrophysiological remodeling [6–8]. Left atrial low-voltage areas (LVAs) identified through electroanatomical mapping are widely considered electrophysiological markers of atrial fibrosis and atrial cardiomyopathy [9, 10]. Moreover, the presence of LVAs has been linked to an increased risk of post-ablation atrial tachyarrhythmia recurrence in patients with PerAF [11–13]. Recent randomized trials have shown that adding LVA modification to pulmonary vein isolation (PVI) does not significantly reduce postablation recurrence in patients with PerAF [12, 14, 15]. Therefore, in addition to intraoperative LVA modification, attention should also be directed toward clinical factors associated with LVA presence. Identifying these factors may improve the understanding of atrial substrate remodeling and provide potential insights for future comprehensive management strategies in patients with PerAF. Previous studies have investigated clinical predictors of LVAs; however, many included mixed populations of paroxysmal and persistent AF, limiting the understanding of factors specifically associated with advanced atrial remodeling in PerAF [8, 11]. In PerAF-specific cohorts, previous studies have primarily focused on demographic characteristics, comorbidities, renal function, and structural remodeling parameters associated with LVAs [16–18]. However, the potential contribution of routinely measured metabolic and lipid-related factors to atrial low-voltage substrates remains insufficiently characterized. Therefore, the present study aimed to identify clinical factors associated with LVAs in patients with PerAF, with particular emphasis on the association between routinely measured metabolic parameters, including lipid profiles, glucose-related parameters, thyroid-related markers, and other related indicators, and atrial low-voltage substrates. Methods Study design and population This single-center retrospective cohort study enrolled consecutive patients with PerAF who underwent first-time catheter ablation at The First People’s Hospital of Changzhou between January 2021 and August 2024. Patients with previous ablation, valvular atrial fibrillation, no intraprocedural left atrial substrate mapping for LVA assessment, or complete loss to follow-up were excluded. Data collection and definitions Clinical, laboratory, electrocardiographic, echocardiographic, and procedural information were retrospectively retrieved from the electronic medical record system of The First People’s Hospital of Changzhou. Baseline variables included demographics, comorbidities, medication use, laboratory results, and relevant imaging and electrophysiological data. Laboratory measurements were obtained from fasting venous blood samples. Serum lipid levels, fasting blood glucose levels, renal function indices, and serum potassium levels were measured using standard automated assays. Postoperative high-sensitivity C-reactive protein (hs-CRP) was available in some patients and was analyzed only in post hoc models rather than as a primary baseline covariate because it reflects postprocedural inflammatory status and was unavailable for many patients. LVAs were defined as contiguous regions with more than three adjacent mapping points, each spaced approximately 0.5 cm apart, with bipolar voltage amplitudes < 0.5 mV [19, 20]. LVA status was determined from operator-documented intraprocedural substrate assessment before substrate modification and was checked against available archived electroanatomical maps. Low-voltage areas caused by ablation lesions, such as pulmonary vein isolation, empirical linear ablation, or vein of Marshall ethanol infusion, were excluded from native LVA classification. Because the mapping rhythm may influence voltage assessment, its potential impact was evaluated in sensitivity analyses. Supplementary quantitative analyses were performed in patients with complete sinus-rhythm maps suitable for standardized measurement, including the number of accepted mapping points per patient, total LVA burden, and regional LVA burden. The detailed procedures are described in the Supplementary Materials. Descriptions of other measurements, including fibrillatory wave amplitude, are provided in the Supplementary Materials. Ablation procedure The center participated in the STABLE-SR-III study [20], and approximately 80% of patients underwent the STABLE-SR ablation strategy. Under CARTO 3 guidance (Biosense Webster, Irvine, CA, USA), a three-dimensional left atrial geometry was created with a PentaRay catheter (Biosense Webster, Irvine, CA, USA). Circumferential pulmonary vein antral isolation was then performed with a SmartTouch SurroundFlow catheter (Biosense Webster, Irvine, CA, USA) until complete electrical isolation was achieved. If AF persisted after PVI, synchronized electrical cardioversion (200 J, biphasic) was performed under intravenous sedation to restore sinus rhythm. When sinus rhythm was restored, substrate mapping was performed. In patients with LVAs, individualized linear ablation or localized substrate modification was performed; in those with an electrically healthy atrial substrate, PVI alone was performed. If sinus rhythm could not be restored, atrial-arrhythmia mapping and subsequent substrate-guided ablation were performed at the operator’s discretion. In selected patients with long-standing persistent AF and left atrial enlargement, adjunctive vein of Marshall ethanol infusion was performed. The detailed ablation procedures and procedural parameters are available in the Supplementary Materials (Supplementary Figures S1 and S2). Postablation management and follow-up Oral anticoagulation was continued for at least 3 months after ablation. After catheter ablation, antiarrhythmic drugs, most commonly amiodarone or propafenone, were routinely prescribed for 3 months to reduce early atrial arrhythmia recurrence. After 3 months, these drugs were generally discontinued in patients who remained in sinus rhythm but could be continued or reintroduced in those with recurrent or intermittent atrial tachyarrhythmia during follow-up. Patients were routinely followed at 1, 3, 6, and 12 months after ablation and annually thereafter if no recurrence occurred. Cardiac rhythm was assessed at each visit by 12-lead electrocardiography or 24-hour Holter monitoring. Endpoint events were further identified through standardized telephone follow-up conducted in December 2025 and a review of outpatient and inpatient records. Outcome measurements The primary effectiveness endpoint was atrial arrhythmia recurrence after 3 months following the index ablation. Recurrence was defined as any episode of AF, atrial flutter, or atrial tachycardia lasting > 30 s after the blanking period. Secondary endpoints included hospitalization or emergency department visits for acute coronary syndrome, ischemic stroke, heart failure, and all-cause mortality. Statistical analysis Data normality was assessed using the Shapiro–Wilk test. Continuous variables were summarized as the mean ± standard deviation or median (interquartile range), as appropriate, and were compared using Student’s t test or the Wilcoxon rank-sum test. Categorical variables were presented as numbers and percentages and were compared using the chi-square test or Fisher’s exact test. Missing hs-CRP values were handled using multiple imputation; details are provided in the Supplementary Methods. Baseline characteristics were first compared between patients with and without LVAs. Candidate factors associated with LVA presence were identified using univariable logistic regression together with consideration of clinical relevance and prior evidence, and parsimonious multivariable logistic regression was subsequently used to assess independent associations. Among the routinely measured metabolic variables, LDL-C was selected for focused analyses because it showed the most consistent association with LVA presence across the initial univariable and multivariable analyses. Incrementally adjusted logistic regression models were constructed to evaluate the association between LDL-C and LVA presence. Models 1a–3a constituted the primary confounder-adjusted analyses. Model 4a additionally included left atrial diameter (LAD) as a prespecified attenuation analysis, whereas Models 5a–6a examined additional metabolic and inflammatory covariates. Exact model specifications are provided in Table 3 and the Supplementary Methods. Table 3Incrementally adjusted logistic regression models of continuous LDL-C levels and left atrial low-voltage areasModelNumber of predictorsOR (95% CI)P valueModel 1a30.683 (0.514–0.908)0.009Model 2a110.672 (0.501–0.902)0.008Model 3a150.652 (0.475–0.895)0.008*Model 4a160.732 (0.529–1.013)0.060Model 5a80.681 (0.508–0.914)0.011**Model 6a90.680 (0.505–0.915)0.011ORs with 95% CIs were estimated per 1 mmol/L increase in LDL-C using logistic regression across incrementally adjusted modelsAbbreviations: LDL-C Low-density lipoprotein cholesterol, LVAs Left atrial low-voltage areas, OR Odds ratio, CI Confidence interval, BMI Body mass index, T3 Triiodothyronine, COPD Chronic obstructive pulmonary disease, HDL-C High-density lipoprotein cholesterol, LAD Left atrial diameter, hs-CRP high-sensitivity C-reactive proteinModel 1a: Adjusted for age and sexModel 2a: Adjusted for age, sex, BMI, T3, hypertension, coronary artery disease, heart failure, diabetes, stroke, and COPDModel 3a: Adjusted for age, sex, BMI, T3, hypertension, coronary artery disease, heart failure, diabetes, stroke, COPD, hyperlipidemia, triglycerides, HDL-C, and statin useModel 4a: Adjusted for age, sex, BMI, T3, hypertension, coronary artery disease, heart failure, diabetes, stroke, COPD, hyperlipidemia, triglycerides, HDL-C, statin use, and LADModel 5a: Adjusted for age, sex, BMI, fatty liver disease, blood glucose, serum albumin, and uric acidModel 6a: Adjusted for age, sex, BMI, fatty liver disease, blood glucose, serum albumin, uric acid, and hs-CRP*Model 4a additionally included LAD as a prespecified attenuation model. Because LAD may represent a downstream structural correlate of the observed association between LDL-C and the atrial substrate phenotype, Model 4a was not included in the primary confounder-adjusted model sequence**Model 6a: hs-CRP values were imputed using multiple imputation (40 datasets), and the model was constructed using data pooled across these imputed datasets. To address potential skewness in hs-CRP values, a logarithmic transformation was applied (i.e., log(hs-CRP + 0.1)) Incrementally adjusted logistic regression models of continuous LDL-C levels and left atrial low-voltage areas ORs with 95% CIs were estimated per 1 mmol/L increase in LDL-C using logistic regression across incrementally adjusted models Abbreviations: LDL-C Low-density lipoprotein cholesterol, LVAs Left atrial low-voltage areas, OR Odds ratio, CI Confidence interval, BMI Body mass index, T3 Triiodothyronine, COPD Chronic obstructive pulmonary disease, HDL-C High-density lipoprotein cholesterol, LAD Left atrial diameter, hs-CRP high-sensitivity C-reactive protein Model 1a: Adjusted for age and sex Model 2a: Adjusted for age, sex, BMI, T3, hypertension, coronary artery disease, heart failure, diabetes, stroke, and COPD Model 3a: Adjusted for age, sex, BMI, T3, hypertension, coronary artery disease, heart failure, diabetes, stroke, COPD, hyperlipidemia, triglycerides, HDL-C, and statin use Model 4a: Adjusted for age, sex, BMI, T3, hypertension, coronary artery disease, heart failure, diabetes, stroke, COPD, hyperlipidemia, triglycerides, HDL-C, statin use, and LAD Model 5a: Adjusted for age, sex, BMI, fatty liver disease, blood glucose, serum albumin, and uric acid Model 6a: Adjusted for age, sex, BMI, fatty liver disease, blood glucose, serum albumin, uric acid, and hs-CRP *Model 4a additionally included LAD as a prespecified attenuation model. Because LAD may represent a downstream structural correlate of the observed association between LDL-C and the atrial substrate phenotype, Model 4a was not included in the primary confounder-adjusted model sequence **Model 6a: hs-CRP values were imputed using multiple imputation (40 datasets), and the model was constructed using data pooled across these imputed datasets. To address potential skewness in hs-CRP values, a logarithmic transformation was applied (i.e., log(hs-CRP + 0.1)) Secondary analyses were conducted to further characterize the LDL-C–LVA association. Restricted cubic splines were used to assess possible nonlinearity, quartile-based analyses provided descriptive comparisons across LDL-C categories, subgroup analyses with false-discovery-rate-adjusted interaction tests assessed potential heterogeneity, and mediation analyses examined whether selected clinical variables might partly account for the observed association. Postablation recurrence was evaluated using Kaplan–Meier analysis and Cox proportional hazards regression. These secondary analyses were considered exploratory and hypothesis-generating, and their detailed methods are provided in the Supplementary Methods. Sensitivity analyses included additional adjustment for mapping rhythm, repetition of the analyses in the sinus-rhythm mapping subgroup, exclusion of patients with coronary artery disease (CAD), and quantitative assessment of regional and total LVA burden in patients with complete sinus-rhythm maps suitable for standardized measurement. Odds ratios (ORs) and hazard ratios (HRs) are reported with 95% confidence intervals (CIs). All tests were two-sided, and P < 0.05 was considered statistically significant. Analyses were carried out in R (version 4.4.1; R Foundation for Statistical Computing, Vienna, Austria). Results Baseline characteristics From January 2021 to August 2024, 454 consecutive patients with PerAF underwent catheter ablation at The First People’s Hospital of Changzhou. After the application of the predefined exclusion criteria, 404 patients were included in the final analysis, including 232 (57.4%) with LVAs and 172 (42.6%) without LVAs (Fig. 1). The baseline characteristics are presented in Table 1. Patient selection flowchart. A total of 454 consecutive patients with persistent atrial fibrillation who underwent catheter ablation were screened. After exclusion of patients with previous atrial fibrillation ablation, valvular atrial fibrillation, loss to follow-up, or no intraprocedural left atrial substrate mapping, 404 patients were included in the final analysis. Among them, 232 patients had LVAs, and 172 did not. Abbreviations: LVAs, left atrial low-voltage areas Table 1Baseline characteristics according to left atrial low-voltage area statusVariable*LVA Absent(n = 172)LVA Present(n = 232)P valueAge, years66.00 [57.75–71.00]68.00 [63.00–74.00]< 0.001Male, n (%)128 (74.4%)123 (53.0%)< 0.001Female, n (%)44 (25.6%)109 (47.0%)< 0.001Smoking history, n (%)38 (22.1%)47 (20.3%)0.746Drinking history, n (%)40 (23.3%)33 (14.2%)0.028Systolic blood pressure, mm Hg129.65 ± 14.56130.15 ± 15.390.739Diastolic blood pressure, mm Hg82.00 [74.00–90.00]83.00 [74.00–89.00]0.975Body mass index, kg/m²25.37 [23.71–28.35]25.00 [23.09–27.65]0.107CHA2DS2-VASc score3.00 [2.00–4.00]4.00 [3.00–5.00]< 0.001Atrial fibrillation duration, years1.00 [0.24–3.00]1.00 [0.24–4.00]0.31 Laboratory findings Total cholesterol, mmol/L4.21 [3.58–4.96]4.00 [3.45–4.67]0.023 Triglycerides, mmol/L1.23 [0.92–1.84]1.23 [0.89–1.74]0.706 HDL-C, mmol/L1.13 [1.00–1.27]1.15 [0.99–1.37]0.531 LDL-C, mmol/L2.52 ± 0.792.30 ± 0.700.003 Fasting blood glucose, mmol/L5.41 [4.95–6.29]5.42 [4.87–6.56]0.977 TSH, µIU/mL1.85 [1.23–2.71]2.34 [1.43–3.17]0.016 T3, pmol/L4.71 [4.30–5.13]4.65 [4.20–4.94]0.029 T4, pmol/L17.06 [15.23–18.43]17.06 [15.41–18.62]0.59 Serum potassium, mmol/L3.97 [3.73–4.19]3.95 [3.64–4.16]0.239 Blood uric acid, µmol/L368.75 [312.47–441.22]363.15 [299.57–436.38]0.473 Blood urea nitrogen, mmol/L5.79 [4.88–6.96]5.78 [4.79–6.82]0.689 Blood creatinine, µmol/L73.00 [62.00–85.20]70.55 [62.00–82.00]0.223 eGFR, mL/min/1.73 m²98.44 [83.48–119.75]96.43 [85.16–113.42]0.34 Serum albumin, g/L39.50 [37.20–42.02]39.20 [37.10–41.42]0.153 **Postablation hs-CRP, mg/L5.55 [2.80–14.25]9.65 [3.46–25.73]0.013 Other auxiliary examinations Left atrial diameter, mm45.00 [42.00–48.00]46.00 [43.75–49.25]< 0.001 Left ventricular ejection fraction, %57.00 [54.00–61.00]58.00 [55.00–61.00]0.405 f-wave amplitude, mV0.15 [0.12–0.20]0.12 [0.08–0.16]< 0.001 Basal heart rate, bpm91.00 [78.00–103.25]88.00 [75.00–104.00]0.319 Cardiovascular history Hypertension, n (%)120 (69.8%)160 (69.0%)0.949 Coronary artery disease, n (%)28 (16.3%)40 (17.2%)0.904 Heart failure, n (%)46 (26.7%)78 (33.6%)0.17 Previous stroke, n (%)28 (16.3%)38 (16.4%)1 Cardiomyopathy, n (%)1 (0.6%)7 (3.0%)0.146 Other medical history COPD, n (%)13 (7.6%)21 (9.1%)0.724 Diabetes, n (%)25 (14.5%)43 (18.5%)0.353 Fatty liver, n (%)51 (29.7%)60 (25.9%)0.465 Renal disease, n (%)11 (6.4%)8 (3.4%)0.252 Hyperthyroidism, n (%)1 (0.6%)3 (1.3%)0.64 Hyperlipidemia, n (%)20 (11.6%)25 (10.8%)0.913 Medications Statins, n (%)58 (33.7%)86 (37.1%)0.555 Antiplatelet, n (%)15 (8.7%)15 (6.5%)0.507 Diuretic, n (%)59 (34.3%)96 (41.4%)0.086 Anticoagulant, n (%)172 (100.0%)231 (99.6%)1 β-blocker, n (%)111 (64.5%)151 (65.1%)0.993 ACEI or ARB, n (%)80 (46.5%)115 (49.6%)0.591 CCBs, n (%)69 (40.1%)106 (45.7%)0.309Continuous variables are presented as the mean ± standard deviation or median [interquartile range], as appropriate. Categorical variables are presented as n (%). Comparisons between groups were performed using Student’s t test or the Wilcoxon rank-sum test for continuous variables, and the chi-square test or Fisher’s exact test for categorical variables, as appropriateAbbreviations: LVAs Left atrial low-voltage areas, HDL-C High-density lipoprotein cholesterol, LDL-C Low-density lipoprotein cholesterol, TSH Thyroid-stimulating hormone, T3 Triiodothyronine, T4 Thyroxine, eGFR estimated glomerular filtration rate, COPD Chronic obstructive pulmonary disease, ACEI Angiotensin-converting enzyme inhibitor, ARB Angiotensin receptor blocker, CCB Calcium channel blocker, LAD Left atrial diameter, hs-CRP high-sensitivity C-reactive protein*LVA status for the present analysis was adjudicated from operator-documented substrate findings before substrate modification and verified against available intra-procedural substrate maps, with lesion-related low-voltage changes excluded**High-sensitivity C-reactive protein was measured postoperatively. In the overall cohort, 66.1% (267 patients) had hs-CRP data available. Specifically, 71.5% of patients in the LVA absent group and 62.1% in the LVA present group had hs-CRP data Baseline characteristics according to left atrial low-voltage area status Continuous variables are presented as the mean ± standard deviation or median [interquartile range], as appropriate. Categorical variables are presented as n (%). Comparisons between groups were performed using Student’s t test or the Wilcoxon rank-sum test for continuous variables, and the chi-square test or Fisher’s exact test for categorical variables, as appropriate Abbreviations: LVAs Left atrial low-voltage areas, HDL-C High-density lipoprotein cholesterol, LDL-C Low-density lipoprotein cholesterol, TSH Thyroid-stimulating hormone, T3 Triiodothyronine, T4 Thyroxine, eGFR estimated glomerular filtration rate, COPD Chronic obstructive pulmonary disease, ACEI Angiotensin-converting enzyme inhibitor, ARB Angiotensin receptor blocker, CCB Calcium channel blocker, LAD Left atrial diameter, hs-CRP high-sensitivity C-reactive protein *LVA status for the present analysis was adjudicated from operator-documented substrate findings before substrate modification and verified against available intra-procedural substrate maps, with lesion-related low-voltage changes excluded **High-sensitivity C-reactive protein was measured postoperatively. In the overall cohort, 66.1% (267 patients) had hs-CRP data available. Specifically, 71.5% of patients in the LVA absent group and 62.1% in the LVA present group had hs-CRP data Compared with those without LVAs, patients with LVAs were older (68.00 [63.00–74.00] vs. 66.00 [57.75–71.00] years, P < 0.001) and more frequently female (47.0% vs. 25.6%, P < 0.001). They also had lower LDL-C levels (2.30 ± 0.70 vs. 2.52 ± 0.79 mmol/L, P = 0.003), a larger LAD (46.00 [43.75–49.25] vs. 45.00 [42.00–48.00] mm, P < 0.001), and a smaller fibrillatory wave amplitude (0.12 [0.08–0.16] vs. 0.15 [0.12–0.20] mV, P < 0.001). Lower total cholesterol (TC) and triiodothyronine (T3) levels, and higher thyroid-stimulating hormone (TSH) and hs-CRP levels, were also observed in patients with LVAs (all P < 0.05). Univariable and multivariable logistic regression analyses Univariable logistic regression revealed that older age, a higher CHA2DS2-VASc score, female sex, a larger LAD, lower LDL-C levels, lower T3 levels, and a smaller fibrillatory wave amplitude were associated with LVA presence, whereas TC levels, TSH levels, and diuretic use were not (Table 2). In the multivariable model, age (OR 1.044, 95% CI 1.009–1.081; P = 0.012), female sex (OR 2.167, 95% CI 1.279–3.670; P = 0.004), LAD (OR 1.077, 95% CI 1.029–1.126; P = 0.001), and LDL-C levels (OR 0.718, 95% CI 0.531–0.971; P = 0.031) remained independently associated with LVAs (Table 2). Table 2Univariable and multivariable logistic regression analyses for factors associated with left atrial low-voltage areasVariablesUnivariable logistic regressionMultivariable logistic regressionOR (95% CI)P valueOR (95% CI)P valueAge, years1.054 (1.029–1.080)< 0.0011.044 (1.009–1.081)0.012CHA2DS2-VASc score1.275 (1.126–1.444)< 0.0010.95 (0.796–1.134)0.569Total cholesterol, mmol/L0.976 (0.876–1.086)0.654––LDL-C, mmol/L0.664 (0.507–0.870)0.0030.718 (0.531–0.971)0.031TSH, µIU/mL1.012 (0.961–1.065)0.659––T3, pmol/L0.69 (0.516–0.924)0.0130.92 (0.670–1.263)0.605Left atrial diameter, mm1.066 (1.023–1.111)0.0021.077 (1.029–1.126)0.001Body mass index, kg/m²0.958 (0.906–1.013)0.137––Hypertension0.963 (0.626–1.475)0.863––Diabetes1.338 (0.787–2.316)0.289––*f-wave amplitude, mV0.000 (0.000–0.005)< 0.0010.001 (0.000–0.024)< 0.001 Sex Malerefref Female2.578 (1.679–3.958)< 0.0012.167 (1.279–3.670)0.004 Drinking history Norefref Yes0.547 (0.328–0.912)0.0210.983 (0.547–1.767)0.955 Diuretic use Noref Yes1.352 (0.898–2.035)0.149––ORs with 95% CIs were estimated using logistic regression. “ref” indicates the reference category; “–” indicates not included in the multivariable modelAbbreviations: OR Odds ratio, CI Confidence interval, LDL-C Low-density lipoprotein cholesterol, TSH Thyroid-stimulating hormone, T3 Triiodothyronine*For f-wave amplitude, estimates may be unstable because of quasi-complete or complete separation; penalized logistic regression results are provided in the Supplementary Materials Univariable and multivariable logistic regression analyses for factors associated with left atrial low-voltage areas ORs with 95% CIs were estimated using logistic regression. “ref” indicates the reference category; “–” indicates not included in the multivariable model Abbreviations: OR Odds ratio, CI Confidence interval, LDL-C Low-density lipoprotein cholesterol, TSH Thyroid-stimulating hormone, T3 Triiodothyronine *For f-wave amplitude, estimates may be unstable because of quasi-complete or complete separation; penalized logistic regression results are provided in the Supplementary Materials Because the fibrillatory wave amplitude showed an extremely strong inverse association with LVAs, conventional logistic regression produced complete separation. Therefore, Firth-penalized logistic regression was applied, indicating that each 0.1-mV reduction in fibrillatory wave amplitude was associated with higher odds of LVAs (OR 2.09, 95% CI 1.44–3.09; P < 0.001) (Supplementary Table S1, Supplementary Materials). Associations between LDL-C levels and LVAs across multivariable models Among the routinely measured metabolic variables evaluated, LDL-C showed the most consistent association with LVA presence across the initial univariable and multivariable analyses. Specifically, LDL-C remained independently associated with LVA presence after multivariable adjustment, whereas the other lipid-related measures, including TC, triglycerides, and HDL-C, and the other metabolic variables did not show the same consistency across the initial analyses. LDL-C was therefore selected for subsequent focused analyses. When analyzed as a continuous variable, higher LDL-C levels were consistently associated with a lower likelihood of LVAs across the primary confounder-adjusted models (Models 1a–3a) (Table 3). After adjustment for age and sex, the OR was 0.683 (95% CI 0.514–0.908; P = 0.009). This association remained similar after further adjustment for clinical comorbidities (Model 2a: OR 0.672, 95% CI 0.501–0.902; P = 0.008) and lipid-related factors, including statin use (Model 3a: OR 0.652, 95% CI 0.475–0.895; P = 0.008). When LAD, a prespecified structural correlate, was additionally entered into the attenuation model (Model 4a), the observed association was weakened and no longer statistically significant (OR 0.732, 95% CI 0.529–1.013; P = 0.060). Additional models adjusting for metabolic status and inflammatory markers continued to show an inverse association between LDL-C and LVAs (P < 0.05). Variance inflation factor analysis revealed no substantial multicollinearity among the covariates (Supplementary Table S2). Subgroup and interaction analyses In subgroup analyses, the relationship between LDL-C levels and LVA presence was generally consistent across clinically relevant strata, although the strength of the association varied (Fig. 2). In the unadjusted interaction analysis, LAD showed nominal heterogeneity (P for interaction = 0.011). The inverse association was stronger in patients with smaller left atria (LAD < 45 mm; OR 0.44, 95% CI 0.27–0.67) and weaker in those with a larger LAD (OR 0.92, 95% CI 0.64–1.32). However, after false discovery rate correction, no interaction remained statistically significant (Supplementary Table S3). Subgroup analyses of associations between LDL-C levels and left atrial low-voltage areas. ORs and 95% CIs were estimated using univariable logistic regression models. P values for interaction were obtained from unadjusted interaction analyses. The association was broadly directionally consistent across subgroups; nominal heterogeneity was observed for LAD. Abbreviations: LDL-C, low-density lipoprotein cholesterol; LVAs, left atrial low-voltage areas; OR, odds ratio; CI, confidence interval; BMI, body mass index; LAD, left atrial diameter; eGFR, estimated glomerular filtration rate Secondary mediation analysis The results of the mediation analyses of all the candidate pathways are shown in Supplementary Figure S3. Among the examined variables, only LAD showed evidence consistent with a possible partial indirect association between LDL-C levels and LVA presence (Fig. 3). Lower LDL-C levels were associated with a larger LAD (path a = − 0.804), and a larger LAD was related to a higher likelihood of LVAs (path b = 0.075). The indirect effect through LAD was significant (average causal mediation effect [ACME] = − 0.013, P = 0.014) and accounted for approximately 15% of the total effect (total effect = − 0.086, P = 0.006). The direct effect also remained significant (average direct effect [ADE] = − 0.072, P = 0.020) (Supplementary Table S4). These findings are compatible with the possibility that the observed LDL-C–LVA association partly reflects left atrial structural remodeling. Mediation analysis of associations among LDL-C levels, left atrial diameter, and left atrial low-voltage areas. LAD was the only examined variable showing evidence consistent with a possible partial indirect association between LDL-C levels and the presence of LVAs. The indirect effect through LAD was significant (ACME = -0.013, P = 0.014), and the direct effect remained significant (ADE = -0.072, P = 0.020). Significance estimates were obtained using 1,000 bootstrap iterations. Abbreviations: LDL-C, low-density lipoprotein cholesterol; LAD, left atrial diameter; LVAs, left atrial low-voltage areas; ACME, average causal mediation effect; ADE, average direct effect Secondary categorical analyses of LDL-C quartiles in relation to LVAs Restricted cubic spline analysis revealed no significant nonlinearity in the relationship between LDL-C levels and LVA presence after adjustment for age and sex (P for nonlinearity = 0.191) (Fig. 4). Although the spline curve showed some visual variation across the LDL-C distribution, this should not be overinterpreted in the absence of formal statistical evidence of nonlinearity. Quartile-based analyses were therefore retained only as descriptive categorizations of LDL-C levels. Restricted cubic spline analysis of LDL-C levels and left atrial low-voltage area presence. Restricted cubic spline analysis based on logistic regression was used to evaluate the association between LDL-C levels and the presence of LVAs, with the median LDL-C level (2.35 mmol/L) as the reference. The model was adjusted for age and sex. Histogram bars indicate the distribution of LDL-C values. No significant evidence of nonlinearity was observed (P for nonlinearity = 0.191), although some visual variation across the LDL-C distribution was apparent. Abbreviations: LDL-C, low-density lipoprotein cholesterol; LVAs, left atrial low-voltage areas Patients were stratified by LDL-C quartiles as follows: Q1, 0.55–1.88 mmol/L (n = 102); Q2, 1.88–2.35 mmol/L (n = 101); Q3, 2.35–2.90 mmol/L (n = 100); and Q4, 2.90–4.82 mmol/L (n = 101). The prevalence of LVAs differed significantly across quartiles and was lower overall in the higher LDL-C quartiles (Q1, 64.7%; Q2, 64.4%; Q3, 52.0%; Q4, 48.5%; P = 0.034) (Supplementary Table S5). In the quartile-based models, Q2 was used as the model reference category for numerical comparison only and should not be interpreted as indicating a biologically or clinically preferred LDL-C range. In the subsequent multivariable logistic regression analyses, both Q3 and Q4 were associated with a lower likelihood of LVAs than Q2 across the three progressively adjusted models. In the age- and sex-adjusted model, the odds ratios were 0.54 (95% CI 0.30–0.97; P = 0.039) for Q3 and 0.53 (95% CI 0.29–0.95; P = 0.034) for Q4, whereas Q1 was not significantly associated with LVA presence. These associations persisted after further adjustment for clinical comorbidities and lipid-related factors, including statin use (all P < 0.05) (Supplementary Table S6). Secondary categorical analyses of LDL-C quartiles and postablation outcomes To evaluate whether differences in LDL-C categories were accompanied by variations in clinical outcomes, post-ablation recurrence was analyzed using the same quartile stratification. Kaplan–Meier analysis revealed differences in recurrence-free survival among the four quartiles (log-rank P = 0.021), with clear separation of the crude recurrence curves across categories (Fig. 5). Kaplan-Meier curves for recurrence-free survival after catheter ablation according to LDL-C quartiles. Kaplan–Meier analysis demonstrated significant differences in recurrence-free survival after catheter ablation across LDL-C quartiles (P = 0.021 by log-rank test), with visible separation of the crude recurrence curves across categories. Numbers at risk are shown below the x-axis. Abbreviations: LDL-C, low-density lipoprotein cholesterol; Q, quartile In the Cox regression analyses, Q3 was used as the model reference category for numerical comparison only; this choice should not be interpreted as indicating a biologically or clinically preferred LDL-C range (Supplementary Figure S4). In the univariable model, compared with Q3, Q2 and Q4 showed higher risks of recurrence (Q2: HR 1.82, 95% CI 1.10–3.02, P = 0.021; Q4: HR 2.07, 95% CI 1.26–3.42, P = 0.004), whereas Q1 was not. The associations for Q2 and Q4 remained significant after adjustment for age and sex and were similar after further adjustment for clinical and lipid-related factors, including statin use. In fully adjusted models incorporating procedural and substrate-related variables, as well as LAD, the association for Q2 was attenuated and became non-significant, whereas the association for Q4 remained statistically significant across models. These findings persisted after additional adjustment for post-blanking antiarrhythmic drug use (Supplementary Table S7). No significant association was identified for Q1 in any model. No obvious violation of the proportional hazards assumption or problematic multicollinearity was detected in any Cox model (Supplementary Tables S8 and S9). In the quartile-based analyses, no significant differences were observed among the LDL-C quartiles with respect to the secondary endpoints (all P > 0.05) (Supplementary Table S10). Sensitivity analyses Sensitivity analyses addressing mapping-rhythm heterogeneity Compared with sinus-rhythm mapping, atrial-arrhythmia mapping was associated with a higher prevalence of LVAs (64 [83.1%] vs. 168 [51.4%], P < 0.001; Supplementary Table S11). Prespecified sensitivity analyses were therefore performed. First, the mapping rhythm was added to the multivariable logistic regression models in the overall cohort. The inverse association between LDL-C and LVA presence remained significant in Models 1c–3c and was attenuated only after further adjustment for LAD in Model 4c, while the direction of the association remained unchanged (Supplementary Table S12). Second, the analyses were repeated in the subgroup restricted to patients who underwent sinus-rhythm mapping only. In the sinus-rhythm mapping subgroup (n = 327, 80.9%), LVAs were identified in 168 patients (51.4%). Patients with LVAs had lower LDL-C levels than those without LVAs (2.30 ± 0.70 vs. 2.54 ± 0.81 mmol/L, P = 0.005; Supplementary Table S13). In progressively adjusted logistic regression models, higher LDL-C levels remained inversely associated with LVA presence across all the models, including the model further adjusted for LAD, although the association was modestly attenuated after LAD was added (Supplementary Table S14). Mediation analysis in this subgroup again revealed a significant indirect effect through LAD (ACME = − 0.016, P = 0.034), accounting for approximately 18% of the total effect (Supplementary Table S15; Supplementary Figure S5). The restricted cubic spline curve in the sinus-rhythm mapping subgroup was similar to that in the overall cohort (Supplementary Figure S6). When the same LDL-C quartile cutoffs were used, no significant differences in recurrence-free survival across quartiles were observed in Kaplan–Meier analysis (log-rank P = 0.098), although a broadly similar qualitative separation of the quartile curves was observed (Supplementary Figure S7). Additional sensitivity analysis excluding patients with coronary artery disease As an additional sensitivity analysis, analyses were performed after excluding patients with CAD (n = 336; Supplementary Table S16). In this restricted cohort, the inverse association between continuously modeled LDL-C and LVA presence remained significant across Models 1e–3e (OR range 0.637–0.658; all P < 0.05) and was attenuated after additional adjustment for LAD in Model 4e (OR 0.719, 95% CI 0.490–1.057; P = 0.093) (Supplementary Table S17). Mediation analysis again revealed a partial indirect pathway through LAD (ACME = − 0.013, P = 0.032; total effect = − 0.098, P = 0.014) (Supplementary Figure S8). Significant differences in recurrence-free survival across LDL-C quartiles were observed in Kaplan–Meier analysis, showing a curve separation pattern similar to that seen in the main cohort (log-rank P = 0.031; Supplementary Figure S9). Supplementary sensitivity analyses of quantitative LVA burden Supplementary analyses showed clear differences according to the mapping rhythm. Compared with sinus-rhythm mapping, atrial-arrhythmia mapping resulted in a lower proportion of complete quantitative maps but a markedly greater number of accepted mapping points and a substantially greater total LVA burden (all P < 0.001; Supplementary Table S18). In contrast, within the sinus-rhythm subgroup with complete quantitative data, patients with and without LVAs had similar mapping completeness and similar numbers of accepted mapping points (Supplementary Table S19). In the sinus-rhythm subgroup with complete quantitative data, LVAs were identified in 157 patients (51.3%), and the posterior wall was the most frequently involved region (34.0%) (Supplementary Table S20). Among patients with LVAs, the median total LVA burden was 9.00% [4.20%–14.70%], and the greatest regional burden was observed in the posterior wall (3.10% [0.00%–5.40%]) (Supplementary Table S20). In the sinus-rhythm subgroup with complete quantitative data, LDL-C levels were inversely associated with total LVA burden in both unadjusted and progressively adjusted models (Supplementary Table S21). In regional analyses, significant inverse associations were observed for inferior wall burden and right pulmonary vein antrum burden, whereas no significant associations were found for anterior wall, posterior wall, or roof burden. Scatter plots also revealed modest inverse correlations between LDL-C levels and total LVA burden both in the overall sinus-rhythm complete-case subgroup and in the subgroup with only LVA presence (R = -0.24, P = 2e-05; and R = -0.26, P = 0.00085, respectively) (Supplementary Figure S10). An additional analysis in the sinus-rhythm subgroup with complete quantitative data revealed that LDL-C levels differed significantly across the five total LVA burden categories and showed an overall decreasing trend with increasing LVA burden, with the lowest LDL-C levels observed in the > 15% burden group (Supplementary Table S22; Supplementary Figure S11). Older age, female sex, and larger LAD were also associated with greater LVA burden across these categories (Supplementary Table S22). Discussion In this study, higher LDL-C levels were associated with a lower likelihood of LVAs in the primary confounder-adjusted models. After the LAD was added as a possible structural correlate, this association became weaker. Attenuation and mediation analyses suggested that left atrial structural remodeling may partly align with the observed association. A broadly similar pattern was observed after excluding patients with CAD, further supporting the robustness of the main findings, although treatment-related confounding cannot be fully excluded. Quartile-based analyses revealed heterogeneity in LVA prevalence and recurrence across LDL-C categories, but these secondary findings were descriptive and should be considered hypothesis-generating. In the present study, age, female sex, LAD, f-wave amplitude, and LDL-C levels remained independently associated with LVA presence after multivariable adjustment. Consistent with previous studies in patients with PerAF, age, female sex, and LAD are well-established correlates of LVAs [16, 17]. Notably, the f-wave amplitude showed a strong inverse association with LVA presence, with smaller amplitudes indicating a greater likelihood of abnormal atrial substrate. These findings are in line with prior electrophysiological and electrocardiographic evidence suggesting that the f-wave amplitude reflects viable atrial myocardium and is inversely related to atrial remodeling and low-voltage substrates [21–23]. Among the routinely measured metabolic variables evaluated in this study, LDL-C showed the most consistent independent association with LVA presence across the primary models. However, given the observational design, this finding does not establish a causal relationship between LDL-C levels and atrial remodeling. Previous research has linked LDL-C levels mainly to AF incidence and postablation recurrence, whereas its relationship with the underlying atrial electrophysiological substrate has been less well studied. Numerous cohort studies and systematic reviews have reported an inverse relationship between LDL-C levels and AF risk, a phenomenon commonly described as the “cholesterol paradox” [24–29]. This association has been described as independent of statin use in some prior studies [25, 29], which is broadly consistent with the stratified analyses and suggests that LDL-C levels, rather than lipid-lowering therapy alone, may be associated with the atrial substrate phenotype. Lower LDL-C levels have also been associated with a higher risk of postablation recurrence [30–32], with some studies suggesting that this pattern may be more pronounced in women [30, 32]. This finding is in line with the observation that female patients had a greater burden of low-voltage substrates. Several possible explanations have been proposed for the “cholesterol paradox”. First, cholesterol is an essential component of cardiomyocyte membranes, and lower membrane cholesterol may disrupt ion channel function and calcium handling, thereby increasing atrial electrical instability and susceptibility to AF [33–35]. Second, low LDL-C levels may reflect an underlying inflammatory or catabolic state rather than a truly protective profile. Systemic inflammation can lower circulating cholesterol levels while also promoting atrial structural remodeling and fibrosis [33, 36]. Third, the risk of AF may be determined by both LDL-C concentration and the quantitative and qualitative properties of lipoprotein particles, including small, dense LDL particles and oxidized LDL [28]. In the present study, the “cholesterol paradox” was further explored at the atrial substrate level by showing an inverse relationship between LDL-C levels and the presence of LVA. Supplementary quantitative analyses in patients with complete sinus-rhythm maps further supported this pattern by revealing lower LDL-C levels in association with greater total LVA burden. This finding is consistent with the possibility that the paradox may not only reflect a statistical link between LDL-C levels and AF onset or recurrence but may also be associated, at least in part, with differences in atrial myocardial structure and electrophysiological characteristics. These findings indicate that LDL-C may be a clinically relevant marker associated with the atrial cardiomyopathy phenotype, with left atrial enlargement as one possible correlate. Mediation analyses suggested that LAD accounted for approximately 15%–18% of the association between LDL-C and LVA presence, indicating that structural remodeling may be one possible, but incomplete, explanatory factor. This is biologically plausible because atrial enlargement reflects chronic hemodynamic stress, myocyte stretch, and extracellular matrix remodeling, all of which can promote fibrosis and low-voltage substrate formation [37, 38]. Low LDL-C levels may also be a marker of unfavorable metabolic or inflammatory states associated with atrial dilation and remodeling, which may in turn be associated with a greater LVA burden [33]. Moreover, the limited proportion of the association accounted for by LAD suggests that atrial size alone does not fully capture the structural or molecular correlates linking LDL-C levels to the atrial substrate. In subgroup analyses, the inverse association between LDL-C levels and LVAs appeared more pronounced in patients with smaller atria; however, this was attenuated after correction for multiple testing and should therefore be interpreted with caution. Overall, these findings are compatible with an association involving atrial structural remodeling, but the underlying mechanisms remain unclear. When LDL-C levels were categorized into quartiles for descriptive purposes, LVA prevalence and recurrence patterns were not uniform across categories; however, these findings should not be interpreted as evidence of a true range-dependent effect. In these descriptive categorizations, LVA prevalence was higher in the lower LDL-C quartiles, whereas recurrence curves showed heterogeneity across quartiles without establishing a biologically meaningful nadir. These quartile-based findings should be interpreted descriptively because there was no formal evidence of nonlinearity for LVAs. In patients without CAD, these results should not be taken as evidence against guideline-directed LDL-C lowering. Rather, they should be viewed as a hypothesis-generating observation on atrial substrate and recurrence patterns in this specific cohort. Any apparent excess recurrence in the highest LDL-C level category may have reflected adverse metabolic influences not captured by LDL-C alone, including a coexisting cardiovascular burden and other aspects of metabolic health [39, 40]. This may help explain why postablation recurrence did not fully parallel the low-voltage substrate burden, suggesting that recurrence may also be influenced by nonsubstrate mechanisms beyond atrial voltage-defined remodeling, such as triggers or the broader metabolic milieu [41]. Mapping-rhythm heterogeneity is an important methodological issue in this study. However, after additional adjustment for mapping rhythm, the inverse association between LDL-C and LVAs remained evident in the overall cohort and was attenuated only after further adjustment for LAD. In the sinus-rhythm mapping subgroup, this inverse association remained significant across all progressively adjusted models, including the model that additionally included LAD. Together, these findings provide directionally consistent supportive evidence for the main results. From a clinical perspective, these findings should not be interpreted as supporting intentional elevation of LDL-C levels or relaxation of lipid-lowering therapy among patients with atrial fibrillation. Given the retrospective observational design, the present results indicate an association rather than a causal or therapeutic effect. Instead, lower LDL-C levels may identify patients at higher risk of advanced atrial remodeling and may therefore be used primarily for risk stratification rather than to guide treatment decisions. Prospective studies with serial lipid measurements are needed before any substrate-oriented implications for lipid management can be considered. Study strengths and limitations This study has several strengths. It includes a relatively large cohort of 404 consecutive patients with PerAF who underwent initial catheter ablation, enhancing statistical robustness. LVAs were assessed using standardized high-resolution electroanatomical mapping with strict quality-control procedures, including the exclusion of lesion-related signals and the use of artifact-reduction strategies, thereby improving measurement reliability and internal validity. In addition, a comprehensive analytical framework was applied, including multivariable logistic regression with incremental adjustment, restricted cubic spline analysis, mediation analysis, subgroup analyses, and sensitivity analyses. Supplementary quantitative analyses of LVA burden and regional distribution in a sinus-rhythm subset further supported the primary findings and strengthened the substrate-level interpretation. Taken together, these strengths support the robustness and interpretability of the observed associations; however, several important limitations should be considered when interpreting the findings. First, its retrospective observational design precludes causal inference regarding LDL-C levels, LAD, and LVA presence. Second, LVA assessment was subject to methodological heterogeneity because voltage mapping was performed during sinus rhythm in most patients but during atrial arrhythmia in a minority. Since bipolar voltage may be lower during atrial arrhythmia, some misclassification of LVA status is possible despite adjustment for mapping rhythm and sensitivity analyses restricted to sinus-rhythm mapping. Third, although supplementary quantitative analyses of total and regional LVA burden were performed in patients with complete sinus-rhythm maps, these data were not available for all patients, and not all left atrial regions were suitable for quantitative assessment. Therefore, the main findings in the full cohort still relied primarily on binary LVA assessment, and the quantitative analyses should be interpreted as supportive rather than confirmatory. Fourth, ablation strategies were individualized mainly according to left atrial substrate characteristics within the STABLE-SR workflow. This procedural heterogeneity may have influenced postablation recurrence outcomes; therefore, recurrence analyses should be interpreted with caution and should not be viewed as reflecting substrate-related risk alone. Fifth, although multivariable adjustment and sensitivity analyses, including the exclusion of patients with CAD, were performed, residual confounding cannot be excluded, particularly because the inflammatory status was incompletely characterized and detailed information on the intensity, duration, and adherence to lipid-lowering therapy was unavailable; hs-CRP was assessed postoperatively rather than preoperatively and only in a subset of patients, so the hs-CRP-adjusted model should be regarded as a post hoc analysis rather than a definitive adjustment for baseline inflammation. Sixth, LDL-C levels were measured at a single preprocedural time point, and longitudinal changes during follow-up were not systematically assessed. Finally, because the study included only patients with PerAF who underwent first-time catheter ablation, the findings may not be generalizable to other AF populations or broader clinical settings. In addition, no formal nonlinear association between LDL-C levels and LVA presence was demonstrated by restricted cubic spline analysis; therefore, the quartile-based analyses should be viewed only as descriptive categorizations of a continuous exposure rather than evidence of a biologically meaningful threshold or preferred LDL-C range. Conclusion In patients with PerAF who underwent first-time catheter ablation, higher LDL-C levels were associated with a lower likelihood of LVAs, suggesting a potential link between lipid metabolism and atrial structural remodeling. Secondary quartile-based analyses of LDL-C demonstrated heterogeneous patterns for LVAs and postablation recurrence. Clinically, lower LDL-C levels may have potential value as a readily available marker for identifying patients with a higher likelihood of advanced atrial substrate remodeling and for supporting preprocedural risk assessment. Because of the retrospective observational design, the observed association does not establish causality. Whether LDL-C has a causal role in atrial structural remodeling remains uncertain and warrants further investigation. Supplementary Information Supplementary Material 1.