The presence of AF was identified with an electrocardiogram (ECG). The inclusion criteria were as follows: (1) patients aged 18 to 74 years; (2) non-valvular AF; (3) CHA₂D-VA score of 0 or 1; and 4) patients undergoing 3D electroanatomic high-density bipolar voltage mapping of the left atrium (LA). The exclusion criteria were as follows: (1) previous AF ablation history; (2) patients with thrombi in the LA; (3) left atrial size ≥55 mm (2-dimensional transthoracic echocardiography); (4) patients with severe structural cardiac disease (severe heart valve regurgitation, hypertrophic cardiomyopathy, or other severe valvular disease); (5) congenital heart disease; or (6) patients with carotid or cerebral artery stenosis.

The CHA₂DS₂-VASc score incorporates the patient's age, sex, and history of ischemic stroke/transient ischemic attack/systemic embolism, and the presence of congestive heart failure, hypertension, diabetes mellitus, and vascular disease. Since female sex is now considered a risk modifier instead of a risk factor for stroke in NVAF patients (7, 14), a non-gender CHA₂DS₂-VASc score (CHA₂DS₂-VA score) of 0 or 1 was defined as a clinically low risk of TE in our study. AF patients with CHA₂DS₂-VA scores of 0 to 1 before the thromboembolic event (TE group) were analyzed and compared with low-risk AF patients without TE history (the non-TE group). Paroxysmal AF (PAF) was defined as continuous AF for <7 days. Non-PAF was defined as continuous AF for >7 days, which included long-standing persistent AF for >1 year.

Antiarrhythmic medications were withheld for 5 half-lives before the procedure. Transesophageal echocardiography or computed tomography scanning was performed to rule out left atrial thrombi for all patients preceding the procedure. After observing the left atrial geometry, pulmonary vein isolation (PVI) was performed under conscious sedation using intravenous fentanyl and via transseptal access to the left atrium. Assessment of LA surface area and LVA followed a clinical routine and were performed as described previously (15, 16). A sequential high-density 3D bipolar voltage map (at least 300 points per patient) was constructed in all patients during sinus rhythm, guided by a 3D electroanatomic mapping system (CARTO3, Biosense Webster, or Ensite velocity, Abbott) using a circular mapping catheter (Lasso, 4 mm interelectrode spacing, Biosense Webster or Advisor FL, 3 mm, interelectrode spacing, Abbott). Electrical cardioversion was performed when the patients were in AF. Stable contact between the local atrial tissue and the electrodes of the mapping catheter was required. Internal point filter software was used to limit data acquisition. Only mapping sites that were within a distance of 5 mm from the acquired LA shell contributed to the voltage map. LVA was defined as contiguous areas (at least 3 adjacent points) of bipolar voltage < 0.5 mV. Extra care was taken at the sites where LVA was recorded to exactly define the extent of LVA.

The LA was segmented into six areas: anterior wall, posterior wall, right pulmonary vein antrum, roof, septum, and lateral wall (Figure 1). The distribution and surface area of the LVA were calculated. The ratio of the LVA to the LA area was defined as the percentage of the LA area within the isolation line, not including the ablation lesions in the surface area of the LA and anatomical structures such as the PV, LAA and mitral annulus. Among patients with left atrial LVA, the activation time from the SN to the LAA was also recorded by measuring the time from the onset of the P wave in Lead II of the surface ECG to the earliest activation of the LAA during sinus rhythm.

All analyses were performed using SPSS 19.0 statistical software. Normally distributed continuous variables are presented as the mean ± standard deviation (SD) and were compared between groups using Student's t-test. Categorical variables are presented as counts with percentages. The difference between groups was tested using the chi-square test. Fisher's exact test was performed for low expected counts (n <5). All parameters with a significance of p < 0.05 in the univariate analysis were entered into the multiple logistic regression analysis. The multivariate logistic regression method was used to determine the association of the clinical variables with the thromboembolic history. A p-value < 0.05 was considered statistically significant.

Table 2 shows that left atrial LVA was more prevalent in the TE group than in the non-TE group (60.0 vs. 29.2%, p = 0.003). Among patients with left atrial LVA, the ratio of the LVA to LA area was similar between these two groups, but the activation time from SN to LAA was significantly longer in the TE group (77.09 ± 21.09 vs. 57.59 ± 15.19 ms, p < 0.001) (Table 2).

Figure 3 depicts the distribution of LVA in the LA among all of the patients with left atrial LVA. In the TE group, 8 patients (53%) had a documented LVA on the left atrial anterior wall, 5 (33%) on the roof region, 4 (27%) on the right pulmonary vein antrum, 1 (7%) on the lateral wall, 1 (7%) on the posterior wall, and none on the septum. In the non-TE group, 24 patients (23%) had a documented LVA on the left atrial anterior wall, 18 (17%) on the roof region, 21 (20%) on the right pulmonary vein antrum, none on the lateral wall, 62 (59%) on the posterior wall, and 4 (4%) on the septum. The left atrial LVA was more preferentially located at the anterior wall in the TE group than in the non-TE group (53 vs. 23%, p = 0.025). The posterior wall was more frequently affected in the non-TE group than in the TE group (62 vs. 7%, p < 0.001).

In the univariate analysis, the existence of anterior LVA in the LA, female sex, and elevated levels of NT-pro BNP were significantly associated with a TE history. Multivariate analysis demonstrated that LVA at the anterior wall of the LA [OR: 4.17 (95% CI: 1.51 to 11.51); p = 0.006] and being female [OR: 3.40 (95% CI: 1.36 to 8.51); p = 0.009] were associated with a history of TE (Table 3).

It has been speculated that some additional independent risk factors for AF-related TE may not be included in the CHA₂DS₂-VASc score (17). In the general NVAF population, several studies have identified additional risk factors (18–20), such as NT-pro BNP, left ventricular diastolic dysfunction, and impaired kidney function. However, in the low-risk AF population, very few studies have evaluated potential predictors of AF-related TE (7, 8). A retrospective, cross-sectional study showed that old age (≥60 years), high NT-pro BNP (≥300 pg/mL), low creatinine clearance (<50 mL/min), and an enlarged LAD (≥45 mm) were independently associated with stroke in clinically low-risk AF patients (7). Another matched case-control study revealed that smoking was a predictor of ischemic stroke in low-risk AF patients (8). In that study, some smokers in the non-stroke group might have been undetected because the patients without records on smoking were all considered non-smokers. This makes the conclusion of the study less compelling. Considering the significant social impact of stroke in these low-risk patients, it is essential to identify novel risk factors that can predict TE in this cohort.

In animal models, an increased LA size was associated with a higher degree of left atrial interstitial fibrosis (21, 22). A clinical study by Fatema et al. (23) reported that LA volume enlargement was a biomarker for ischemic stroke. However, since left atrial fibrosis may occur before apparent enlargement of the LA, echocardiographic measurement of the LA size may result in an underestimation of the association of LA fibrosis and thromboembolic events.

Daccarett et al. and Akoum et al. (9, 10) elegantly showed that a high degree of left atrial fibrosis, identified by DE-MRI, is independently associated with the left atrial thromboembolic milieu and a prior history of stroke in patients with AF. In addition, recent studies revealed that left atrial LVA (a bipolar voltage range of 0.1–0.5 mV), detected using 3D electroanatomic mapping during sinus rhythm and used as a surrogate marker of left atrial fibrosis, was associated with postprocedural silent cerebral events and a history of stroke (12, 13). These results are consistent with our findings of the association between left atrial LVA and TE. In our study, left atrial LVA was more prevalent in the TE group than in the non-TE group. However, unlike previous studies, the total prevalence of LVA was relatively low in our study. A major reason is that all patients enrolled in our study had few comorbidities. In addition, the catheters used in our study had a large inter-electrode gap, which reduced the accuracy of the maps and may partly explain the different results.

In our study, patients in the TE group were more likely to be women. Pfannmüller et al. (24) reported that fibrosis-related genes were upregulated in postmenopausal women with AF. A histological study analyzing atrial tissue obtained during cardiac surgery illustrated that women had higher expression of CX 40 than men, which indicated remodeling-induced changes in connexins (25). In our study, multivariate analysis demonstrated that being female was associated with LVA at the anterior wall of the LA (Table 4), and LVA at the anterior wall of LA was more prevalent in patients with a prior history of TE. However, based on our analysis, the prevalence of LVA at the posterior wall of the LA did not differ between male and female (15.3 vs. 19.4%, p = 0.352). Meanwhile, the increased LVA at the posterior wall in the non-TE group did not reach the significant difference (p = 0.097). Thus, being prone to fibrosis at the anterior wall of the LA may be an important reason to explain why female sex was a risk factor for TE in clinically low-risk AF patients in our study. However, considering that the association of female sex with higher risk of TE was independent of LVA at the anterior wall of the LA, the specific mechanism of the association between female sex and TE needs further investigation.

It has been recognized that most emboli arise from the LAA in patients with AF (26). In our study, the left atrial LVA was more preferentially located at the anterior wall in the TE group than in the non-TE group. LA fibrosis existing in the anterior wall region has been shown to be correlated with a reduction in LAA flow velocity (27). While this reduction is strongly related to the presence of emboli in patients with AF. In our study, we additionally found that the activation time from the SN to the LAA was significantly longer in the TE group than in the non-TE group. This delayed conduction from the SN to the LAA, which was dominantly determined by the anterior wall fibrosis, could indirectly reflect a reduction in LAA flow velocity. Therefore, our findings partially provide additional plausible evidence for an association among anterior LA fibrosis, LAA function and thromboembolic events.

There is increasing evidence showing that AF may not be a prerequisite for embolic stroke (28, 29). Instead, AF could be considered as a manifestation of fibrotic atrial cardiomyopathy that has progressed to a certain stage (30, 31). The underlying mechanism of embolic stroke might be related to atrial fibrosis and atrial cardiomyopathy. Under such conditions, LA endothelial functional impairment might exist, with a probably abnormal coagulation state and delayed LAA emptying function. All of these factors could work together, and LAA clot formation might be the final consequence.

Based on the results of this study, we propose that for clinically low-risk AF patients, especially for women, LVA at the anterior wall of the LA could be used to refine the risk stratification for the risk of TE and guide the anticoagulation strategy following PVI.

The inherent limitation of this study is its cross-sectional nature. Although we illustrated that LVA in the anterior wall of the LA and female sex were independently associated with a history of TE in clinically low-risk AF patients, further prospective studies are still needed. Second, it should be noted that LA voltage mapping was performed after the TE occurred. The delay between the TE events and the mapping varied. Furthermore, LA remodeling is a dynamic process, which may have an unknown effect on our observed finding. However, we reported here findings of a hypothesis-generating study that illustrated the association between LVA at the anterior wall of the LA and TE. Third, whether LVA detected by 3D electroanatomic voltage mapping could accurately reflect LA fibrosis is still debatable. Fourth, since the number of patients was relatively small, it was difficult to analyze the correlation among LAA 3D geometry, LAA Doppler flow velocity and TE, but this will be performed in our future study. Fifth, it seems to be an overstatement to conclude that the presence of low-voltage areas alone is associated with thrombotic risk. Other data, specifically serological data, information an any abnormal coagulation status and assessment of left atrial and left ventricular function by ultrasound, would also be required. Sixth, both recorded low voltage area and the intra-atrial conduction time were measured in sinus rhythm. Since it is in AF rhythm that thrombosis is actually more likely to occur, more functional mapping of the left atrium would be the preferred method of assessment. The novelty of this paper would be enhanced by the addition of different pacing sites and the evaluation of low potential areas and conduction times during rapid atrial pacing. Finally, this study was a single-center study. The study results might be hard to extrapolate to general NVAF patients with low CHA₂DS₂-VA scores.