Prevalence and multiple predictors of left atrial low voltage in paroxysmal and non-paroxysmal atrial fibrillation patients undergoing ablation: a systematic review and meta-analysis

Abstract

Aims

In the left atrium (LA), low-voltage areas (LVAs) detected at electroanatomic mapping in patients with atrial fibrillation (AF) are considered expression of atrial cardiomyopathy (AtCM). This meta-analysis aims at assessing the prevalence and predictors of LVAs in a larger AF population undergoing catheter ablation.

Methods and results

Studies comparing patients undergoing LA ablation with vs. those without LVAs were included. Meta-analyses were conducted to estimate the prevalence and odds ratios (ORs) for LVAs. Twenty-two studies with 5278 patients were included. Low-voltage areas were present both in paroxysmal (28%) and non-paroxysmal (41%) patients. The strongest predictors of LVA presence were: age > 65 years (OR 3.41), CHA₂DS₂-VASc score (OR 3.29), non-paroxysmal AF (OR 3.19), NT-proBNP > 365 pg/mL (OR 2.47), female sex (OR 2.40), E/e′ ratio (OR 2.31), eGFR < 60 mL/min/m² (OR 2.28), and LA volume indexed > 34 mL/m² (OR 1.98). Comorbidities were also predictors but with lower ORs. In subgroup analysis, female sex (OR 3.90) was a predictor only in non-paroxysmal, while LA diameter (OR 2.51) and body mass index (BMI; OR 1.85) positively correlated only in paroxysmal AF. Meta-regression analysis showed that non-paroxysmal AF and age were independently and significantly associated with a greater reduction in BMI in patients with compared to those without LVAs.

Conclusion

Low-voltage areas can be present in both paroxysmal and non-paroxysmal AF, and can be predicted by multiple clinical, echocardiographic, and biomarker variables. The impact of female sex, LA diameter, and BMI on LVA presence varies according to the type of AF.

Graphical Abstract

Graphical Abstract.

Introduction

Three decades ago, the term atrial remodelling was introduced to describe the electrical and eventually the structural changes occurring during the persistence of atrial fibrillation (AF).¹ Structural remodelling, including fibrosis, was then reported to play a key role in creating a substrate initiating and maintaining AF establishing a vicious circle in which stressors such as heart failure (HF) and hypertension contribute to worsen both the arrhythmia and the atrial substrate.² More recently, with the evidence that several patient characteristics and comorbidities concur to structural remodelling, the occurrence of AF has been put in the context of an atrial cardiomyopathy (AtCM), defined as any complex of structural, architectural, contractile, or electrophysiological changes affecting the atria which can potentially produce clinical manifestation.³

At electroanatomic mapping, in patients with AF a colocalization of low-voltage areas (LVAs) in the left atrium (LA), defined as regions with bipolar voltage <0.5 mV in sinus rhythm, with areas of delayed enhancement at cardiac magnetic resonance (CMR)⁴ was found, as well as a correlation between reduced bipolar voltage and atrial fibrosis.⁵ Therefore, LVAs have been considered a surrogate of local fibrosis.⁶ Recently, however, further evidence has been provided that LVAs are characterized by a more altered histology including increased intercellular space, myofibrillar loss, and decreased nuclear density.⁷ The complexity of histologic alterations better characterize the complexity of the dynamic processes involved in AtCM which includes, among others, genetic factors, inflammatory signalling, ageing, and female sex.⁶

Although a recent consensus document states that ablation of LVAs in addition to pulmonary vein isolation (PVI) in persistent AF is an area of uncertainty,⁸ predicting the presence of LVAs can be helpful for accurate preprocedural assessment of ablation candidates, in whom the need for adjunctive ablation and/or continuation of antiarrhythmic drug therapy during follow-up may be required.

Multiple scores exist to predict LVAs,⁹ but they are based on a restricted number of predictors identified in single studies with limited prospective validation. Hence, in candidates for AF ablation, there is the need to evaluate the prevalence and predictors of LVAs in the LA in a larger cohort of patients with subgroup analyses, which represents the aim of this systematic review and meta-analysis.

Methods

Protocol

The study protocol adhered to the recommendations of the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement^10,11 and the Meta-analysis of Observational Studies in Epidemiology (MOOSE) reporting guidelines.¹² The study protocol was registered in the Prospective Register of Systematic Reviews (PROSPERO ID: CRD42024522942). As the present research was a meta-analysis of published studies, the approval from an ethics committee or informed consent from patients was waived. The dataset is available at reasonable request.

We developed a PECO (Participants, Exposure, Comparator, and Outcomes) statement,¹³ which included the following:

• Participants: humans.

• Exposure: age, sex, non-paroxysmal type of AF, E/e′ ratio, absolute LA volume (LAV), LA volume indexed (LAVi), LA anteroposterior diameter (LAD), left ventricular ejection fraction (LVEF), diabetes mellitus (DM), hypertension, HF, CHA₂DS₂-VASc score, body mass index (BMI), estimated glomerular filtration rate (eGFR), B-type natriuretic peptide (BNP), N-terminal prohormone of brain natriuretic peptide (NT-proBNP), months since AF diagnosis (AF duration), history of stroke or TIA.

• Comparators: not applicable.

• Outcomes: any LVAs in the LA on endocardial voltage mapping regardless of the grade of their extension.

Information sources and search strategy

We identified all studies that referenced predictive factors for LA LVAs in patients with AF and were available with an English abstract. Electronic search engines included Medline via PubMed, Web of Science, Scopus, and Cochrane Central. Relevant Medical Subject Headings and keywords were employed and were searched from inception through 31 December 2024. The search terms for each database are reported in Supplementary material online, Table S1. We also performed an additional manual search of secondary sources, such as reference lists of retrieved articles, to comprehensively identify relevant studies (i.e. backward and forward snowballing).

Eligibility criteria and study selection

Studies comparing populations with and without LVAs in patient undergoing PVI for AF were considered eligible for inclusion in the analysis. Only articles published in peer-reviewed journals were included. Published abstracts and meeting presentations were excluded. The search was restricted to articles in the English language. Studies that included patients with a history of AF ablation or prior heart surgery, lacked quantitative presentation of population data or did not include a control group were excluded. The quality of studies was assessed based on standard criteria. The Newcastle–Ottawa scale¹⁴ with scores of 0–9 was used to determine the methodological quality of the included studies.

Two authors (C.O. and M.M.) independently screened titles and abstracts to identify studies that met the inclusion criteria. Case studies, reviews, and animal studies were not included. Duplicate entries identified across various electronic databases were removed, and the version with the largest sample size was retained. The remaining reports underwent independent full-text screening by the same reviewers to confirm compliance with the eligibility criteria and the relevance of the content. Disagreements were solved by consensus under the supervision of the lead investigator (R.D.P.).

Data extraction

Information on the authors, year of publication, and sample size of each study were gathered. Study-level data were extracted and organized into dedicated electronic spreadsheets, including outcomes of interest (presence of any LVAs), primary clinical and procedural characteristics, design features, endpoint definitions, and outcomes. The following main population characteristics were identified and collected: age, sex, DM, hypertension, CHA₂DS₂-VASc score, HF, BMI, eGFR, BNP, NT-proBNP, HF history, AF type (paroxysmal vs. non-paroxysmal), AF duration, previous stroke or TIA, echocardiography-assessed LA parameters (LAV, LAVi, LAD), E/e′ ratio, and LVEF.

Study endpoints

The pre-specified primary endpoint of interest was the presence of any LVAs.

Risk of bias assessment

Each article within the final shortlist was then subjected to a complete quality evaluation. The risk of bias was estimated using the Risk Of Bias In Non-randomized Studies of Exposures (ROBINS-E) tool.¹⁵ Any differences in risk of bias rating were resolved through mutual discussion and consensus among reviewers. Finally, the impact of small-study effects and publication bias was assessed by visually inspecting comparison-adjusted funnel plots.

Statistical analysis

The Mantel–Haenszel method was applied for dichotomous data to calculate the 95% confidence intervals (CIs) and odds ratios (ORs); hence, all results were reported as ORs with corresponding 95% CIs. The inverse variance method was used for continuous data to calculate 95% CIs and aggregated mean difference (MD). A random-effects model was used to pool the data. Continuous data effect sizes were converted from MD into ORs following the method described by Hasselblad to ease the clinical interpretation.¹⁶ The process is reported in detail in the Supplementary  Methods  S1. Likewise, to convert continuous variables to categorical variables above a pre-specified cut-off, we followed Suissa’s method.^17,18 The I² statistic was used to assess heterogeneity among studies, with values between 25 and 50% indicating low heterogeneity, 50–75% indicating moderate heterogeneity, and values above 75% indicating high heterogeneity. In order to estimate the weighted prevalence of certain variables, we performed prevalence meta-analysis as reported in Supplementary Methods  S2. Subsequent prevalence subgroup meta-analysis was performed to further investigate I². In this study, all analyses were conducted using a random-effects model. Sensitivity analysis was performed to address between-study heterogeneity in the meta-analysis. Two methods were implemented. First, random-effects model was used to evaluate the consistency of the overall exposure across different pooling approaches. Second, Baujat plots were used to identify studies that contributed most significantly to heterogeneity, and they were then removed to evaluate the robustness and variability of our estimates, if they were considered to present peculiarities in patient selection, methodology, and/or data reporting. Additionally, subgroup analyses were conducted to investigate potential sources of heterogeneity between patients with paroxysmal vs. non-paroxysmal AF and between Asian and Western studies. In the first case, a sensitivity analysis was conducted, including studies that investigated patients with both paroxysmal and non-paroxysmal AF, in order to assess the changes in heterogeneity across the subgroups. A mixed effects meta-regression model was performed using Knapp–Hartung adjustments to assess the relation of covariates (mean age in LVA cohort, type of AF, and study design) on predicted difference of BMI among cohorts. A test for subgroup difference was conducted using a significance threshold of 0.10, while other P-values were considered significant at 0.05.^19,20 Statistical analysis was conducted using R version 4.4.2 (R Foundation for Statistical Computing, Vienna, Austria) meta and metafor packages, Review Manager software, version 5.4.1 (Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2020), and SPSS (version 29.0, IBM).

Results

Study selection and characteristics

Figure 1 shows the PRISMA flowchart summarizing the search strategy. The initial search resulted in 2064 articles. After removing duplicates and after title and abstract analysis 53 articles remained for assessment in the eligibility phase. Thirty-one were then removed for the reasons specified in Figure 1, and the remaining 22 articles were included in the qualitative and quantitative synthesis.^21–42

Figure 1.

PRISMA flowchart reporting the search strategy.

Table 1 summarizes the key characteristics of the included studies with an overall number of 5278 patients (median 172, and range between 50 and 1004 patients per study). Most patients were male (65%) with a mean age of 64.10 ± 4.70 years. Prevalence meta-analysis (Figure 2) showed an overall prevalence of 35% (95% CI 0.29–0.41). Higher prevalence was observed in studies including only non-paroxysmal AF (41%; 95% CI 0.29–0.54) compared to those with only paroxysmal AF (28%; 95% CI 0.18–0.39). The majority (13) of the studies included were observational, while others (9) were designed as interventional trials to evaluate the efficacy of voltage-guided substrate modification beyond PVI. Common exclusion criteria across the studies encompassed the presence of atrial thrombus, severe valvular heart disease, prior cardiac surgery or prior LA ablation, and inability to maintain sinus rhythm during voltage mapping. In all the studies reported, LVAs were defined as areas with <0.5 mV on bipolar voltage mapping in sinus rhythm. However, there were differences in the surface extension cut-offs in determining the grade of the LVAs. In the included studies, voltage mapping was always performed during stable sinus rhythm and, when possible, before the ablation. Although the mapping catheters used for voltage mapping varied among studies, multipolar or contact force sensing catheters were used in all studies resulting in high resolution mapping in the vast majority of the studies (see Supplementary material online, Table S2). In 13 studies, patients were classified binarily according to the presence or absence of LVAs, whereas in the remaining studies they were subclassified according to the percent extension of LVAs with a threshold ≤5–10% and the most severe grade when more than 20–30% of the LA surface was involved (see Supplementary material online, Table S3).

Table 1.

Characteristics of the included studies

Study	Design/(country)	Mean [median] age, years; (%females)	SS	Definition of outcomes	Inclusion criteria	Exclusion criteria	Predictors of LVA	Newcastle-Ottawa scale
Aksan 202236	Cohort (Turkey)	LVA+: 63.8; (66.7%) LVA−: 57.2; (40.4%)	21 94	Regions where bipolar peak to peak voltage was <0.5 mV	Symptomatic paroxysmal AF undergoing catheter ablation	Prior catheter or surgical ablation of AF, left atrial thrombus, persistent and permanent AF, mechanical prosthetic heart valve(s), moderate or severe mitral stenosis.	Female gender [OR = 7.54 (CI: 1.01–59.19)]; LA volume [OR = 1.33 (CI: 1.05–1.67)]; galectin-3 levels, sampled from CS [OR = 1.7 (CI: 1.17–2.48)]	7
Baskovski 202237	Cross-sectional (Turkey)	LVA+: 58.9; LVA−: 55.7	103 121	Area with <0.5 mV voltage	AF undergoing catheter ablation of AF, with a valid electroanatomical map in SR and a previous ECG in SR	Prior history of AF ablation	P-wave dispersion (scar vs. non-scar group: 46 ms ± 20 vs. 38 ms ± 15, P < 0.001)	7
D'Ambrosio 202027	Cohort (Italy, Germany)	LVA+: [62]; (32%) LVA−: [71]; (69%)	299 75	Confluent areas of ≥1 cm2 as determined by at least three adjacent points with bipolar peak-to-peak voltage amplitudes ≤0.5 mV	Symptomatic, drug-refractory, paroxysmal and persistent AF undergoing catheter ablation	Long-standing persistent AF, the inability to obtain sinus rhythm before or after PVI, previous PVI procedure, age <18 years, any cardiac surgery during the prior 3 months, presence of an intracavitary thrombus, severe mitral regurgitation, and missing preprocedural cardiac CT	Female gender [OR = 4.71 (2.52–8.83)]; LA volume index ≥ 57 mL/m2 [OR = 5.48 (CI: 2.73–10.98)]; Age ≥ 65 years [OR = 3.23 (CI: 1.68–6.20)]	7
Hojo 202132	Cross-sectional (Japan)	LVA+: 73; (50%) LVA−: 65.9; (23.7%)	30 118	Voltage values <0.5 mV >5% of the total LA surface area without pulmonary veins	AF undergoing catheter ablation of AF, with a valid electroanatomical map	Patients without a sufficient voltage map, with a history of cardiac surgery or percutaneous coronary intervention	Coronary artery calcifications [OR 1.13 (CI: 1.02–1.24]	7
Ikoma 202238	Cross-sectional (Japan)	LVA+: 67.9; (33.3%) LVA−: 63 (18.2%)	51 308	Area of > 5 cm2  with a bipolar electrogram amplitude of < 0.50 mV	AF undergoing catheter ablation of AF, with a valid electroanatomical map in SR	Age under 18, immediate recurrence of AF during mapping	Age [OR = 1.06 (CI: 1.02–1.10)]; female gender [OR = 2.98 (CI: 1.34–6.61)]; DCM [OR = 8.34 (CI: 1.38–50.37)]; HCM [OR = 5.04 (CI: 1.31–19.36)]; persistent AF [OR 4.19 [1.93–9.10)]; LA volume > 113 mL [OR = 3.21 [1.38–7.50)]	8
Kiedrowicz 202239	Cohort (Poland)	LVA+: [65]; (29%) LVA-: [61] (8%)	65 51	Electrogram amplitude < 0.5 mV. The extent of global LVA burden was arbitrarily considered as mild (< 5% of the TSA), moderate (5–20% of the TSA) and severe (> 20% of the TSA)	Long-standing persistent AF undergoing catheter ablation	Previous ablation for AF, cardiac surgery affecting the atria, severe valvular disease or mechanical valve, known pulmonary hypertension, history of myocarditis or pericarditis	Female gender [OR = 2.41 (CI: 1.34–4.48)]; valvular heart disease [OR = 1.86 (CI: 1.20–2.88)]; CHA2DS2-VASc ≥ 4 [OR = 3.16 (CI: 1.09–9.14)]	8
Kornej 201822	Cross-sectional (bioAF cohort = 241, Leipzig cohort = 215) (Germany/UK)	bioAF-LVA+: [69]; (85%) bioAF-LVA−: [63]; (20%) Leipzig-LVA+: [69]; (62%) Leipzig-LVA−: [60]; (12%)	65 176 58 157	Electrogram amplitude <0.5 mV	Paroxysmal and persistent AF undergoing catheter ablation	Pregnancy, age <18 or >75, valvular AF, cancer, acute or systemic inflammatory diseases	APPLE score [OR = 1.79 (CI: 1.32–2.43)]; DR-FLASH score [OR = 2.14 [1.59–2.89)]	8
Matsuda 202028	Cross-sectional (Japan)	LVA+: 73; (56%) LVA−: 67; (29%)	206 798	Peak-to-peak bipolar voltage of <0.5 mV covering ≥5 cm2 of the left atrium	Paroxysmal and persistent AF undergoing catheter ablation	Patients who could not undergo voltage mapping, decompensated heart failure who could not undergo catheter ablation, age <20 years old, left atrial thrombus, prior catheter ablation of AF	Age ≥70 years [OR = 2.3 (CI 1.5− 3.4)]; female gender [OR = 3.4 (CI: 2.2–5.2)]; persistent atrial fibrillation [OR 1.8 [1.1–3.0)]; diabetes [OR = 1.8 (CI: 1.1–2.8)]; elevated BNP ≥100 pg/dL [OR = 1.7 [1.02–2.8)]	8
Moreno-Weidmann 202133	Cohort (Spain, Germany, Switzerland)	LVA+: 66; (34%) LVA−: 59; (22%)	123 169	Bipolar voltage <0.5 mV in SR with an extension area greater than 2 cm2	Paroxysmal and persistent AF undergoing catheter ablation	Prior LA ablation or surgery, inability to restore sinus rhythm at the beginning of the procedure or to maintain it during the acquisition of the LA-voltage map or any contraindication for AF ablation	Age [OR = 1.06 (CI: 1.01–1.1)]; amplified P-wave duration (145–160 ms vs. <145, OR = 13.2, P = 0.004; > 160 vs. <145, OR = 41.0, P < 0.001) systolic pulmonary artery pressure > 35 mmHg [OR = 3.0 (CI: 1.4–6.4)]; left atrial volume index >38 mL/m2 [OR = 2.5 (CI: 1.2–5.4)]	8
Nakatani 202029	Cross-sectional (Japan)	LVA+: 66; (35%) LVA−: 64; (33%)	26 24	Bipolar voltage of <0.5 mV. The significant LVA was defined as total LVA > 10%	Paroxysmal AF undergoing catheter ablation	Previous catheter ablation, prior heart surgery, thyroid diseases, and pulmonary diseases	P-wave amplitude measured in ECG leads correlates with the presence of local LVAs.	7
Nery 201823	Cohort (Canada)	LVA+: 63.9 (25.6%) LVA−: 59 (26.7%)	43 61	At least three adjacent points showing atrial voltage ≤0.5 mV	Paroxysmal and persistent AF undergoing catheter ablation	Not stated	Age [OR = 1.05 (CI: 1.00–1.11)], LA volume [OR = 1.02 (CI: 1.01–1.04)]	7
Nordin 202442	Cross-sectional (Sweden)	LVA+: 70.3 (54.0%) LVA−: 63.6 (16.9%)	113 326	Area of ≥3 cm2 exhibiting a peak-to-peak bipolar voltage of <0.5 mV.	Non-paroxysmal AF undergoing catheter ablation	LA diameter >55 mm, acute coronary syndrome, severe aortic or mitral valvular heart disease, adult congenital heart disease, previous percutaneous or surgical AF ablation or atrioventricular junction ablation, previous surgery that included left or right atrium, survival <1 year, contraindication for oral anticoagulants (OAC), AF attributed to a reversible condition, pregnancy, or inability or unwillingness of the patient to provide informed consent.	Female gender [OR = 5.78 (CI: 3.61–9.25)]; Age [OR = 1.13 (CI: 1.1–1.17)]; BMI [OR = 0.94 (CI: 0.88–1.0)]; CAD [OR = 3.42 (CI: 1.63–7.17)]; Hypertension [OR = 2.11 (CI: 1.33–3.33)]; LAD [OR = 1.05 (CI: 1.01–1.1)]; NT-proBNP [OR = 1.07 (CI: 1.04–1.1)];	8
Omuro 202134	Cross-sectional (Japan)	LVA+: 66 (27.1%) LVA−: 60.9 (10.9%)	107 46	Points <0.5 mV and covering >5% of the LA body surface area	Non-paroxysmal AF undergoing catheter ablation	Age <20 years, prior surgery of the heart, lungs, or oesophagus, radiotherapy due to cancer in the thorax or previously receiving chemotherapy, prior catheter ablation.	Female gender [OR = 4.02 (CI: 1.05–15.42)]; echocardiographic findings, namely estimated PAPs [OR = 1.12 (CI: 1.19–1.24)]; arrhythmogenic SVC [OR = 3.13 (CI: 1.19–8.23)]	7
Ooie 202030	Cross-sectional (Japan)	LVA+: 73.2 (84%) LVA−: 66.7 (36%)	35 92	Bipolar peak-to-peak voltage amplitude of <0.5 mV and covering >5% of the LA surface area	Paroxysmal and persistent AF undergoing catheter ablation	Treatment with Class I or III antiarrhythmic agents, abnormal electrolyte, implantation of pacemaker or ICD, unavailability of LA-voltage mapping under sinus rhythm before PVI	P-wave duration [OR = 1.12 (CI: 0.87–1.1)]; advanced inter-atrial block [OR = 372.2 (CI: 11.89–11645.17)]	8
Schade 202135	Cohort (Germany)	LVA+: 72 (62%) LVA−: 65 (30%)	65 54	>3 adjacent voltage points of <0.5 mV	Persistent AF with ≥ 2 episodes and inefficacy, contra-indications or rejection of antiarrhythmic medication by the patient	Paroxysmal AF, persistent LA thrombus, and previous catheter or surgical ablation for AF	Age > 67 years [OR = 4.4 (CI: 1.4–13.7)]; LAVI ≥ 68 mL/m2 [OR 3.9 (CI: 1.4–10.5)]; GFR ≤ 85 mL/min/1.73 m2 [OR 12.5 (CI: 2.0–76.6)]; BMI ≥26 kg/m2 [OR = 0.06 (CI: 0.01–0.30]	8
Seewoster 201824	Cross-sectional (Germany/ Netherlands)	LVA+: 69 (57%) LVA−: 63 (35%)	56 160	Electrogram amplitude <0.5 mV	Highly symptomatic paroxysmal and persistent AF, refractory to the antiarrhythmic treatment	Pregnancy, aged <18 or >75 years, valvular AF, cancer, acute or systemic inflammatory diseases	Age [OR = 1.09 (CI: 1.03–1.15)]; female gender [OR = 2.61 (CI: 1.01–6.76)]; LA volume assessed by CMR [OR = 1.02 (CI: 1.01–1.03)]; LA volume assessed by echocardiography [OR = 1.03 (CI: 1.00–1.05)]	8
Shao 202240	Cross-sectional (China)	LVA+: [65] (40.4%) LVA−: [60] (30.9%)	52 162	≥3 adjacent low-voltage points < 0.5 mV	Paroxysmal and persistent AF undergoing catheter ablation	Incomplete for CTA image or contraindication; Sinus rhythm cannot be restored after radiofrequency catheter ablation; Previous history of radiofrequency catheter ablation; History of rheumatic valvular disease, moderate to severe valvular stenosis or insufficiency, congenital heart disease; Cardiac dysfunction; Severe hepatic and renal insufficiency, thyroid dysfunction, respiratory disease, and history of malignant tumour	Age [OR = 1.04 (CI: 1.00–1.08)]; LAVI [OR = 1.02 (CI: 1.00–1.04)]; LA-epicardial adipose tissue volume [OR = 1.19 (CI: 1.00–1.04)]; LA-epicardial adipose tissue attenuation [OR = 0.80 (CI: 0.70–0.92)]	7
Wang 201825	Cohort (China)	Mild LVA+: 64.1 (37%) Moderate LVA+: 66.1 (54.6%) Severe LVA+: 70.1 (88.9%) LVA−: 63 (23.2%)	54 22 18 56	Electrogram amplitude <0.5 mV	Paroxysmal and symptomatic AF undergoing catheter ablation	Left atrial thrombus detected by transoesophageal echocardiography, moderate to severe valves stenosis or insufficiency, contraindication for anticoagulation, systolic heart failure with LVEF <45%, prior catheter or surgical ablation of AF, unable to provide informed consent	Female gender [OR = 3.63 (CI: 1.70–7.75)]; LA surface area [OR = 1.20 (CI: 1.00–1.05)].	7
Wang 202031	Cohort (China)	LVA >20%: 62.4 (30.8%) LVA <20%: 56.4 (30.4%)	52 125	Points with bipolar voltage <0.5 mV covering >20% of the left atrium surface	Paroxysmal and symptomatic AF in pts refractory or intolerant to antiarrhythmic drugs, undergoing catheter ablation	LA appendage thrombosis; abnormal cardiac structures disease (i.e. severe mitral, tricuspid, aortic malformation); HF; history of RFCA, cryoballoon ablation, and cardiac surgical procedures; mental disease; estimated glomerular filtration rate (eGFR) < 30 mL/min; septic shock; advanced malignant tumour; pregnancy; cardiac tamponade or major hydropericardium; or LA anteroposterior diameter >50 mm.	Serum soluble ST2 [OR = 1.61 (CI: 1.38–1.88)]	8
Wang 202241	Cross-sectional (China)	LVA+: 66.7 (44%) LVA−: 65.3 (37.3%)	91 110	Bipolar peak-to-peak voltage amplitude of <0.5 mV and covering >5% of the LA surface area	Paroxysmal and persistent AF undergoing catheter ablation	Prior cardiac surgery, prior AF catheter ablation, severe valvular heart disease, and incomplete imaging data	CHA2DS2-VASc score [OR = 1.26 (CI: 1.02–1.56)]; paroxysmal AF [OR = 2.64 (CI: 1.41–4.92)]; LA diameter [OR = 1.05 (CI: 1.01–1.10)]; total EATVi [OR = 1.01 (CI: 1.01–1.04)]	8
Yamaguchi 201826	Cohort (Japan)	LVA Stage I (<5%): 62 (18%) LVA Stage II (5–20%): 63 (33%) LVA Stage III (20–30%):72 (79%) LVA Stage IV (≥ 30%): 69 (50%)	113 27 14 18	Area with bipolar electrograms <0.5 mV (Stage I < 5%, II 5–20%, III 20–30%, IV >30%)	Non-paroxysmal AF undergoing LA-voltage based catheter ablation	Prior AF ablation, those with LA diameter in the parasternal long-axis view >60 mm, those with severe valvular heart disease, and those who failed SR restoration by external biphasic direct-current cardioversion	Predictors of LVA Stage III/IV: Age [OR = 1.11 (CI: 1.04–1.18)]; Female sex [OR = 4.24 (CI:1.76–10.57)]; Underlying heart disease [OR = 1.96 (CI: 1.06–8.19)]; CHA2DS2-VASc score > 4 [OR = 3.82 (CI: 1.41–10.29)]; LA volume/BSA [OR = 1.03 (CI: 1.01–1.05)]a	8
Yin 201721	Cross-sectional (China)	LVA+: 58.5 (46.7%) LVA−: 54.2 (52.2%)	45 23	>3 adjacent low-voltage points where peak-to-peak bipolar voltage amplitude was <0.5 mV	Persistent AF undergoing catheter ablation	Previous catheter ablation or cardiac surgery	Increase of atrial natriuretic peptide during exercise [OR = 1.63 (CI: 1.03–2.47)]	7

Studies are sorted by alphabetical order based on the first author's name.

AF, atrial fibrillation; APPLE score, Age > 65 years, Persistent AF, imPaired eGFR (<60 mL/min/1.73 m²), left atrial diameter ≥43 mm, left ventricular ejection fraction < 50%; DR-FLASH score, diabetes, renal dysfunction, persistent form of atrial fibrillation, LA diameter >45 mm, age >65 years, female sex, hypertension; BSA, body surface area; CI, 95% confidential interval; CS, coronary sinus; CTA, computed tomography angiography; EATVi, epicardial adipose tissue volume index; eGFR, estimated glomerular filtration rate; LA, left atrium; LVA, low-voltage areas; LVEF, left ventricular ejection fraction; NA, non-retrieved; OR, odds ratio; RFCA, radiofrequency catheter ablation; SR, sinus rhythm; SS, sample size; ST2, suppressor of tumourigenicity 2; SVC, superior vena cava.

^aThe following results were selected among three multivariate regression models presented in the original study, according to the statistical significance of each predictor.

Figure 2.

Forest plot of the prevalence of patients with left atrial low-voltage areas in studies including only patients with paroxysmal atrial fibrillation (A), only patients with non-paroxysmal atrial fibrillation (B), and those including both paroxysmal and non-paroxysmal cases (C).

All studies were rated of high methodological quality on the Newcastle-Ottawa scale (Table 1). According to the ROBINS-E risk of a bias assessment tool for non-randomized observational studies of exposures (see Supplementary material online, Figure S1), studies had a low overall risk of bias. The funnel plot inspection did not suggest publication bias.

Analysis of the outcome

The main findings are summarized in Table 2. Among the identified predictive factors for LVAs, age > 65 years [OR 3.41 (95% CI 2.51–4.65)], CHA₂DS₂-VASc score [OR 3.29 (95% CI 2.70–4.00)], and non-paroxysmal AF [OR 3.19 (95% CI 2.48–4.11)] were the strongest predictors. NT-proBNP > 365 pg/mL, female sex, E/e′ ratio, eGFR < 60 mL/min/m², LAVi > 34 mL/m², and LAD also showed a substantial association with LVAs presence with an OR ranging from 2.47 to 1.97. Comorbidities, including HF, DM, previous stroke/TIA, and hypertension, were all significant predictors, with ORs ranging from 1.38 to 1.61. Figure 3, 4, and 5 show forest plots of the continuous and dichotomous variables, and comorbidities identified as predictors. LVEF was a weak protective factor with a negative association [OR 0.87 (95% CI 0.80–0.94)]. Notably, in the general population, previous stroke or TIA, AF duration >24 months, and BMI were not significantly associated with LVAs presence. Among the considered variables, the LAV showed the highest value of OR [3.49 (95% CI 1.95–6.25)], however, the LAVi with a cut-off value of 34 mL/m² is definitely more prospectively usable. The LAVi and LAD had comparable ORs (1.98 vs. 1.97) and the latter variable, although being a gross indicator of LA enlargement, was considered in much more studies and patients.

Table 2.

Predictors of the low-voltage areas presence in the left atrium at electroanatomic voltage mapping

Factors	No. of participants (No. of studies)	Relative effect (95% CI)
Age > 65 years (vs. age ≤65)	4956 (20)	OR = 3.41 (2.51 to 4.65)
CHA2DS2-VASc score	3642 (12)	MD = 0.94 (0.68 to 1.19) OR = 3.29 (2.70 to 4.00)a
Non-paroxysmal AF (vs. paroxysmal AF)	3571 (11)	OR = 3.19 (2.48 to 4.11)
NT-proBNP >365 (vs. ≤ 365) pg/mL	1802 (3)	OR = 2.47 (1.17 to 5.22)
Female sex (vs. male sex)	5278 (22)	OR = 2.40 (1.92 to 3.00)
E/e′ ratio	1761 (6)	MD = 1.78 (1.01 to 2.54) OR = 2.31 (1.77 to 2.99)a
eGFR < 60 (vs. ≥ 60) mL/min/m2	3782 (13)	OR = 2.28 (1.91 to 2.71)
Left atrial volume indexed > 34 (vs. ≤ 34) mL/m2	1431 (8)	OR = 1.98 (1.38 to 2.84)
Left atrial antero-posterior diameter (mm)	4217 (16)	MD = 2.31 (1.38 to 3.24) OR = 1.97 (1.72 to 2.26)a
Heart failure (vs. no heart failure)	2568 (9)	OR = 1.61 (1.07 to 2.43)
Diabetes mellitus (vs. no diabetes mellitus)	4606 (20)	OR = 1.43 (1.20 to 1.69)
Previous stroke or TIA (vs. no stroke or TIA)	1660 (9)	OR = 1.41 (0.95 to 2.10)
Hypertension (vs. no hypertension)	4662 (20)	OR = 1.38 (1.08 to 1.75)
Atrial fibrillation duration > 24 (vs. ≤ 24) months	2412 (12)	OR = 1.19 (0.93 to 1.52)
BMI (kg/m2)	4204 (16)	MD = −0.02 (−0.67 to 0.62) OR = 1.00 (0.85 to 1.17)a
Left ventricular ejection fraction (%)	4336 (17)	MD = −0.54 (−1.12 to 0.04) OR = 0.87 (0.80 to 0.94)a

BMI, body mass index; CI, confidence interval; eGFR, estimated glomerular filtration rate; MD, mean difference; OR, odds ratio; TIA, transient ischaemic attack.

^aOR in the continuous variable should be interpreted as how one-unit change impacts the odds.

Figure 3.

Forest plot of the following predictors of left atrial low-voltage areas represented by continuous variables: CHA₂DS₂-VASc score (A), E/e′ ratio (B), and anteroposterior left atrial diameter (C).

Figure 4.

Forest plot of the following predictors of left atrial low-voltage areas defined as dichotomous variables: age > 65 years (A), female sex (B), non-paroxysmal atrial fibrillation (C), estimated glomerular filtration rate < 60 mL/min/m² (D), and indexed left atrial volume > 34 mL/m² (E). Note: Kornej(1) = bioAF cohort; Kornej(2) = Leipzig Heart cohort.

Figure 5.

Forest plot of the following comorbidities predicting left atrial low-voltage areas: heart failure (A), hypertension (B), diabetes mellitus (C), and previous stroke/transient ischaemic attack (D).

Three of the studies included follow-up data after substrate modification by ablation. After a 3-month blanking period, the 1-year recurrences of atrial tachyarrhythmia were worse in LVA patients who underwent substrate modification than in LVA-free patients (see Supplementary material online, Figure S2).

3.3. Subgroups analyses: paroxysmal vs. non-paroxysmal atrial fibrillation

Prevalence meta-analysis showed a lower prevalence of female sex in non-paroxysmal than in paroxysmal AF patients (30 vs. 38%, respectively; Supplementary material online, Figure S3). As shown in Table 3, studies focusing exclusively on paroxysmal or non-paroxysmal AF patients revealed differences in ORs. Female sex was strongly associated with LVAs in non-paroxysmal AF [OR 3.90 (95% CI 2.60–5.87)], but not in paroxysmal AF [OR 1.85 (95% CI 0.94–3.65)] (see Supplementary material online, Figure S4A), with subgroup heterogeneity (P = 0.07), suggesting a two-fold stronger association in the non-paroxysmal group. Including studies considering both paroxysmal and non-paroxysmal AF patients [OR 2.18 (95% CI 1.71–2.77)] further reinforces those differences (heterogeneity P = 0.04) (see Supplementary material online, Figure S4B).

Table 3.

Prevalence and effect of the factors predicting left atrial low-voltage areas in paroxysmal vs. non-paroxysmal atrial fibrillation patients

	Paroxysmal AF		Non-paroxysmal AF		P for interactions*
	N studies (prevalence of predictor**)	OR (95% CI)	N studies (prevalence of predictor**)	OR (95% CI)
LAV (mL3)	1	22.65 (13.40–38.28)	2	1.55 (1.20–1.99)	0.15
Age (years)	3	2.76 (2.09–3.65)	5	3.76 (3.16–4.47)	0.62
CHA2DS2-VASc	1	8.17 (4.95–13.46)	3	2.52 (1.94–3.27)	0.07*
Female sex	4 (37%)	1.85 (0.94–3.65)	6 (29%)	3.90 (2.60–5.87)	0.07*
E/e′ ratio	1	1.96 (1.21–3.16)	1	2.10 (1.48–2.98)	0.24
eGFR mL/min/m2	3	0.80 (0.50–1.31)	2	0.35 (0.29–0.42)	0.31
LAVi mL/m2	1	5.31 (3.25–8.66)	4	2.18 (1.81–2.62)	0.93
LAD (mm)	3	2.51 (1.70–3.71)	5	1.13 (0.91–1.40)	0.07*
Heart failure	1 (6%)	0.44 (0.04–5.19)	4 (22%)	1.26 (0.72–2.22)	0.41
Diabetes mellitus	4 (17%)	1.15 (0.64–2.1)	6 (16%)	1.33 (0.83–1.89)	0.69
Previous stroke or TIA	3 (9%)	1.96 (0.85–4.55)	2 (9%)	1.13 (0.58–2.18)	0.31
Hypertension	4 (53%)	0.95 (0.42–2.11)	6 (63%)	1.63 (1.22–2.18)	0.21
AF duration (months)	2	2.43 (0.89–6.61)	3	0.99 (0.83–1.18)	0.44
BMI (kg/m2)	2	1.85 (1.13–3.00)	5	0.57 (0.38–0.86)	0.01*
LVEF (%)	3	0.73 (0.47–1.14)	5	1.04 (0.90–1.20)	0.56

^* P < 0.1 considered statistically significant.

^**For categorical variables.

AF, atrial fibrillation; BMI, body mass index; CHA₂DS₂-VASc, congestive heart failure, hypertension, age ≥75 (doubled), diabetes, stroke (doubled), vascular disease, age 65–74, and sex category (female); E/e′, ratio of early mitral inflow velocity to early diastolic mitral annular velocity; eGFR, estimated glomerular filtration rate; LAD, left atrial diameter—posterior; LAV, left atrial volume; LAVi, left atrial volume indexed; LVAs, low-voltage areas; LVEF, left ventricular ejection fraction; N, number; OR, odds ratio; TIA, transient ischaemic attack.

Hypertension showed a mild trend in non-paroxysmal AF [OR 1.63 (95% CI 1.22–2.18)] but none in paroxysmal AF [OR 0.95 (95% CI 0.42–2.11)], though the subgroup difference was not significant (P = 0.21) (see Supplementary material online, Figure S5A). This was further confirmed when the studies considering both paroxysmal and non-paroxysmal AF patients were introduced in the analysis (heterogeneity P = 0.46) (see Supplementary material online, Figure S5B).

Conversely, LAD was significantly associated with LVAs in paroxysmal AF [OR 2.51 (95% CI 1.70–3.71)], but not in non-paroxysmal AF [OR 1.13 (95% CI 0.91–1.40)], with significant heterogeneity (P = 0.07) and a 2.2-fold stronger association in paroxysmal AF (see Supplementary material online, Figure S6A). Including the studies considering both paroxysmal and non-paroxysmal AF patients [OR 3.17 (95% CI 1.90–4.43)] further confirmed the heterogeneity (P = 0.005) (see Supplementary material online, Figure S6B).

Atrial fibrillation duration did not differ significantly among the subgroups (see Supplementary material online, Figure S7A and B). Of note, only two studies including only paroxysmal AF patients reported this variable.

Body mass index positively correlated with LVAs only in paroxysmal AF [OR 1.85 (95% CI 1.13–3.00)] but had an inverse correlation in non-paroxysmal AF [OR 0.57 (95% CI 0.38–0.86)] with significant heterogeneity (P = 0.01) (see Supplementary material online, Figure S8A). Heterogeneity persisted (P = 0.04) after including studies considering both paroxysmal and non-paroxysmal AF patients [OR 1.15 (95% CI 1.00–1.32)] (see Supplementary material online, Figure S8B).

The CHA₂DS₂-VASc score was associated with a higher OR in paroxysmal AF, but only one study including paroxysmal AF considered this variable. The remaining variables included in this subgroup analysis, such as age, DM, and LVEF, showed a similar correlation with LVAs in the three subgroups of studies.

Subgroups analyses: geographical distribution of the studies

Comparing studies based on the geographic region in which they were conducted, a certain degree of consistency was observed, with an exception. As shown in Supplementary material online, Figure S9, hypertension was significantly associated with LVAs in Western studies only [OR 1.75 (95% CI 1.39–2.20)], while Asian studies showed only a non-significant trend [OR 1.22 (95% CI 0.87–1.71)]. Heterogeneity among the groups was significant (P = 0.08).

Sensitivity analyses

The Baujat plots shown in Figure 6 assess and visualize the impact of between-study heterogeneity and the overall estimated effect contribution in the analysis with high heterogeneity. In studies investigating the association between LAD and LVAs (Figure 6A), heterogeneity was significantly driven by Schade,³⁵ Seewoster,²⁴ D’Ambrosio,²⁷ Baskovski,³⁷ and Kornej²² (BioAF cohort). Excluding these studies from the analysis according to the predefined methodology reduced heterogeneity from an I² of 81 to 53%, while the overall effect estimate remained stable [OR 2.08 (95% CI 1.37–2.79)].

Figure 6.

Baujat plots of the studies investigating the relationship between left atrial low-voltage areas and left atrial diameter (A), CHA₂DS₂-VASc score (B), body mass index (C), and estimated glomerular filtration rate (D). In (A), the arrow indicates the studies of D’Ambrosio²⁷ and Baskovski.³⁷

Similarly, for the association between the CHA₂DS₂-VASc score and LVAs (Figure 6B), the studies by Wang⁴¹ and Kornej²² (Leipzig cohort) stand out as significant outliers, highly contributing to the overall heterogeneity. Their removal decreased the I² from 80 to 3%, with the mean effect remaining significant [random effect mean difference: 0.95 (95% CI 0.83–1.07)].

Regarding the association between BMI and LVAs (Figure 6C), Schade³⁵ and Aksan³⁶ were significant outliers. Their removal from the analysis significantly reduced heterogeneity from an I² 81 to 58%. Overall random effect remained not significant after the removal. Finally, as shown in Figure 6D, studies by Wang³¹ and Shao⁴⁰ were responsible for the heterogeneity in the analyses of eGFR and LVAs. Their exclusion reduced I² from 74 to 0% without modifying the overall effect [random effect mean difference −8.11 (95% CI −9.44; −6.78)]. These findings confirmed the robustness of the results, which are not disproportionately influenced by outliers, which were considered to have peculiarities in patient selection, methodology, and/or data reporting.

Meta-regression

Multiple meta-regression analyses were conducted to explore the interrelated associations between continuous variables, such as eGFR, E/e′ ratio, LAVi, and age. Following exploratory analyses that revealed no significant associations, additional univariate and multivariate meta-regressions were performed to assess the impact of BMI on age in both the overall population and the cohort with LVAs, adjusting for the study design and type of AF. Notably, (Table 4) non-paroxysmal AF (β = −2.86, P = 0.03) and age (β = −0.22 per year, P = 0.02) were independently and significantly associated with a greater reduction in BMI among patients with LVAs compared to those without (Figure 7).

Table 4.

Univariate and multivariate meta-regression analyses assessing the predicted size effect of body mass index, according to age, type of atrial fibrillation, and study design

	Univariate meta-regression		Multivariate meta-regression
Variables	Beta coefficient (95% CI)	P value	Beta coefficient (95% CI)	P value
Non-paroxysmal AF	−2.99 (−5.60 to −0.39)	0.03*	−2.86 (−5.46 to −0.26)	0.03*
Paroxysmal and non-paroxysmal AF	−1.39 (−3.82 to 1.04)	0.24	−1.26 (−4.03 to 1.51)	0.37
Retrospective study design	0.64 (−1.16 to 2.45)	0.46	1.24 (−0.35 to 2.84)	0.11
Mean age in patients with LVAs	−0.18 (−0.38 to 0.02)	0.07	−0.22 (−0.40 to −0.04)	0.02*

*P < 0.05 is statistically significant.

AF, atrial fibrillation; CI, confidence interval; LVAs, low-voltage areas.

Figure 7.

Bubble plot from meta-regression of mean age in the cohort with low-voltage areas vs. predicted difference in body mass index between the two cohorts, with and without low-voltage areas.

Discussion

Prevalence and predictors of low-voltage areas

Patients exhibiting LVAs at LA electroanatomic mapping in sinus rhythm during the first procedure of AF ablation represent about one third of this specific population of AF patients with, however, a wide range of prevalence across studies. This heterogeneity is likely to be due to the different criteria used among centres to select patients for the non-routinely performed procedure of high-density LA mapping in sinus rhythm rather than to heterogeneous criteria to define LVAs. In fact, considering also the increasing availability of concurrent methodologies for AF ablation, the variable criteria to perform LA high-density mapping in sinus rhythm among centres is expected to have a strong impact on the proportion of patients found with LVAs. On the other hand, in all studies, the voltage threshold for LVAs was 0.5 mV, in the vast majority of them adequate technologies were used and high-density mapping obtained, and, although not all studies defined binarily the presence/absence of LVAs, the remaining studies adopted a lower value (≤5–10%) of percent involvement of the LA to identify patients with LVAs, who, in this meta-analysis were all included in the LVA group regardless of the grade of substrate alteration. Interestingly, although the prevalence of patients with LVAs was higher in non-paroxysmal compared to paroxysmal cases, the 95% CI values overlap in these two populations. Consequently, the AF type does not necessarily per se indicate a higher probability of LVAs, as its presence may depend on a variety of other clinical variables.

Moreover, the presence of LVAs in the LA in this specific population of patients can be predicted by multiple variables with ORs ranging from 1.97 to 3.41, including easily accessible clinical characteristics, echocardiographic parameters assessing the LA dimensions and left ventricular filling, and biomarkers. Comorbidities, such as HF, DM, and hypertension, although being significant predictors, show lower ORs, ranging from 1.38 to 1.61. Of note, hypertension shows a significantly different impact on LVAs presence according to the regional distribution of studies, suggesting the possible influence of racial and/or environmental factors.

Finally, our study was not designed to evaluate the impact of LVA ablation for which further evidence is needed. However, as in previous studies,^43–48 the presence of LVAs in the LA is of interest, as it has been reported to be associated with a lower 1-year efficacy of AF ablation.

Controversies in ablation of low-voltage areas and pitfalls in pre-procedure prediction of low-voltage areas

Currently, the efficacy of LVA ablation is controversial in the EHRA Consensus Document this is considered as an area of uncertainty⁸ especially in non-paroxysmal cases: while a randomized trial failed to demonstrate that LA fibrosis ablation guided by CMR has advantage over PVI-only ablation in terms of efficacy and safety outcomes in patients with persistent AF,⁴⁹ a post hoc analysis of the same study,⁵⁰ another randomized trial,⁵¹ and multiple meta-analysis^52–55 showed that ablation of LVAs in addition to PVI was generally associated with a reduction of atrial arrhythmia recurrences in patients with non-paroxysmal AF, possibly with a higher rate of adverse events observed in the randomized trial.⁵¹ In fact, the observation of colocalization of extra-pulmonary vein triggers and LVAs in most patients exhibiting extra-pulmonary vein foci during AF ablation^48,56 provides an intriguing rationale to pursue additional LVA ablation. Interestingly, although further research in this field is necessary, experimental evidence favours the hypothesis of possible uncoupling between cardiomyocytes and fibroblasts via gap junction channels, which may be arrhythmogenic under certain conditions.⁵⁷

Cardiac magnetic resonance has been proposed and widely used to assess atrial fibrosis potentially associated with LVAs⁴ and new software has become available to improve the reproducibility of atrial fibrosis assessment using this non-invasive method.⁵⁸ Cardiac magnetic resonance can also characterize the functional impact of the AtCM, as it detects a significant reduction of the total LA emptying fraction in AF patients with only a mild (9.9%) extension of LVAs in the LA.⁵⁹ However, although a correlation between fibrosis detected at late gadolinium enhancement and LVAs at electroanatomic mapping has been demonstrated,⁴ a large heterogeneity in the study design, patient population, and methodology for CMR acquisition and postprocessing has been observed among studies.⁶⁰ This, in turn, may explain the discordance in localization of the altered LA substrate by CMR compared to electroanatomic mapping.⁶¹ Moreover, differences in the regional distribution of LA fibrosis, reflecting an inhomogeneous process underlying AtCM, limit the value of CMR in predicting clinical outcomes in patients undergoing CMR-guided substrate modification in addition to PVI.⁶² Consequently, for these and other reasons, including CMR availability and added costs, CMR has not become a widespread pre-procedure routine to assess LA fibrosis.

In this scenario, the prediction of LVAs in the large population candidates for AF ablation mainly relies on clinical and echocardiographic variables. Over the years, multiple scores, as the DR-FLASH,⁶³ APPLE,²² ZAQ,²⁷ SPEED,²⁸ and modified APPLE,⁶⁴ have been developed to predict the presence of LVAs in patients undergoing AF ablation. They are based on a limited number of variables, including clinical characteristics, echocardiographic parameters, and comorbidities, and they may become less applicable when parameters derived by a computed tomography scan²⁷ or CMR⁶⁴ are considered. These scores can be used to identify AF patients with higher probability of LVAs in the LA ahead of the ablation procedure with a higher score corresponding to an increasingly higher prevalence of LVAs²⁸ and a lower AF ablation efficacy. However, these scores have been developed in relatively limited cohorts of patients, ranging from 241²² to 1004²⁸ and, when prospectively validated, the validation cohort included several hundreds of patients only in one study.²²

The data from our meta-analysis, including more than 5000 patients, show that in an increased sample size, multiple clinical and echocardiographic variables, or biomarkers can predict the LVA presence. Eight of them have an OR > 2.0, so they can be strong predictors for the population considered. Comorbidities, such as DM and hypertension, included as main determinants in some previously elaborated scores, although they remain significant predictors, have lower ORs. Moreover, subgroup analyses showed that several variables included in predictive scores, including female sex, hypertension, LAD, and BMI, have a substantially different strength in paroxysmal vs. non-paroxysmal AF or in different regional environments. Finally, based on pathophysiologic considerations, all these multiple variables are expected to play a synergic role in determining the development of LVAs in AF patients, but unexpected interactions among these predictors cannot be excluded.

Therefore, based on all these considerations, the development of a simple, universally accepted, and strongly predictive score for LVA presence seems currently a real challenge. While further research possibly involving artificial intelligence is expected to help, the suspected presence of LVAs has to be validated by accurate intraprocedural electroanatomic mapping.

Paroxysmal vs. non-paroxysmal atrial fibrillation

Although it is generally speculated that a greater AF burden in non-paroxysmal compared to paroxysmal AF can be associated to a more extensive presence of LVAs in the LA, the clinical presentation, per se, does not necessarily correlate with the severity of AtCM, as a relatively high prevalence (28%) of patients exhibiting LVAs has been observed in our meta-analysis and other studies including only patients with paroxysmal AF.²⁵ This observation is corroborated by the STABLE-SR-III randomized trial⁶⁵ reporting that patients with an average age of 70 years with paroxysmal AF show LVAs in roughly 30% of the cases and that, in these cases, ablation of LVAs in addition to PVI, results in significantly fewer recurrences in a 23-month follow-up. Interestingly, these data are complemented by the previously published STABLE-SR-II randomized trial,⁶⁶ in which LVAs in the LA were found in only one-half of the cases in a persistent AF cohort of male patients of 60 years of age, on average. Finally, absence of LVAs was found in persistent AF patients with enlarged LA undergoing a successful PVI-only procedure, suggesting that in some non-paroxysmal patients AtCM affects first the atrial size before generating LVAs.⁶⁷ Consequently, in daily clinical practice, additional ablation beyond PVI should be supported by a rationale based on accurate patient evaluation rather than on the mere paroxysmal or non-paroxysmal presentation.

As already mentioned, in our data, some variables show a significantly different strength in predicting LVA presence in the two types of AF. Female sex has a significantly weaker impact in paroxysmal AF, whereas LAD and BMI were significantly associated with LVAs only in paroxysmal AF. This suggests that in paroxysmal AF compared to non-paroxysmal AF, several variables can act differently to generate a certain degree of AtCM regardless of a lower AF burden. Interestingly enough, subgroup and meta-regression analyses indicate that conditions associated with higher BMI may contribute to initiating AtCM specifically in an early stage. However, its evolution may become self-perpetuating and increasingly age-related, mostly driven by cumulative comorbidities independent of BMI and the related metabolic influences.

Limitations

Several limitations should be considered and the interpretation of these results both in the overall population and in the subgroup analyses should be made cautiously, also considering that the OR may not be constant across the entire range of continuous variables as assumed by the logit model.⁶⁸ First, we considered the presence of any LVAs dichotomously rather than considering their burden in the LA. Consequently, the variables identified in this meta-analysis should be used as preprocedural markers of possible LVAs presence useful for shared decision-making and, possibly, procedural shaping and not as predictors of procedural outcomes, which, clearly, may also depend on the severity of the AtCM. Second, this meta-analysis mainly included retrospective studies with varying patient populations, leading to clinical and statistical heterogeneity in some analyses. Third, although the electroanatomic mapping performed in various studies could be defined as ‘high-density’, different catheters, mapping systems, and strategies were used. Moreover, substrate electroanatomic mapping for LVA assessment is rather operator dependent, as well as the definition of the grade of LA involvement. These represent intrinsic limitations that could reduce the comparability of LVAs assessment and increase the heterogeneity among studies. Fourth, dichotomizing continuous variables, especially when their association with outcomes may be continuous and non-linear, is a known constraint of study-level meta-analyses. Fifth, due to the use of aggregate data, we could not formally assess collinearity between covariates. Finally, while this meta-analysis expands on previous smaller cohort studies, its study-level design limits external validity.

Conclusions

The novelty of our meta-analysis including more than five thousand patients resides in two aspects. First, although the prevalence of patients with LVAs is higher in non-paroxysmal AF, the presence of LVAs is not necessarily associated with the AF type. In fact, in paroxysmal AF, the prevalence of LVAs is 28% and this underscores that AtCM can develop even at an early stage of the arrhythmic disease and it may specifically depend on LA enlargement, and, less intuitively, on higher BMI and sex. Second, the multiplicity of predictors, their possible interaction, and different strengths in subgroups render the elaboration of a predictive score challenging and limit its applicability in a wide AF population. This calls for a tailored patient-based approach and for further research aiming at improving non-invasive methodologies and standardize invasive identification of LVAs, possibly in multicentre randomized controlled trials further exploring their presence and impact on long-term clinical outcomes.

Supplementary material

Supplementary material is available at Europace online.

Funding

None.

Data availability

The data supporting the findings of this meta-analysis are derived from previously published studies, all of which are cited in the manuscript. The dataset of this meta-analysis is available at reasonable request.