Atrial fibrillation (AF) is the most common sustainedtachyarrhythmia in the general population and is associated with a high risk of thromboembolism leading to ischemic events ¹-³.

Transthoracic echocardiography (TTE) and transesophageal echocardiography (TEE)are necessary tools to diagnose the presence of left atrialthrombosis (LAT)andLA spontaneous echo contrast (SEC). Both conditions are possibly related to the occurrence of cardiac thromboembolism ³.

Starting from TEE findings of anticoagulated and non-anticoagulated patients referred for electrical cardioversion (ECV) of stable persistent (> 48hrs) non-valvular AF, we investigated the rate of SEC and LAT and identified their independent predictors.

Patients were referred to our tertiary care facility from 1/2012 to 6/2015. Demographic, clinical, and echocardiographic data were collected and stored in the electronic medical recording system. Additional data ⁴ were collected in a separate database for the purpose of performing a doctoral thesis on the specific topic.

The patients included in the present analysis are not consecutive. In fact,onlypatientsthathad been in AF for at least 48 hours before being referred for ECV are included. Furthermore, none of the included patients had been treated for AF within the premises of our facility before. Elective ECV had been planned after having documented a pharmacologically unresponsiveness, defined as failure or intolerance after admission of at least one class I or III anti-arrhythmic drug and/or after patient/referring physician choice. None of the investigated patients presented significant cardiac valve malfunction/disease and none had been previously submitted to cardiac surgical operations.

All patients or their legal representatives signed informed consent to treatment, data collection, and data analysis for scientific research purposes. An ethical commission revision was waved being the study a retrospective analysis of prospectively collected patients´ data and having the study no effect upon the standard of patients´ management.

Data are presented as absolute numbers, percentages, mean ± standard deviation for normally distributed variables and median with 75% IQR values for variables with non-normal distribution. We assessed for normal distribution using the Shapiro-Wilk test. Patients had been referred from different referring physicians and anticoagulation management was not homogenous. We first compared patients´ demographic and clinical data according to the anticoagulation regimen. A second grouping was performed according to the presence of LAT and SEC at pre-ECV TEE. We identified 4 groups (with/without LAT and with/without SEC). Analysis of the variance (ANOVA), Student t-test, Wilcoxon signed-rank test, χ-square, and Fischer-exact test were used whenever appropriate to compare groups (according to anticoagulation regimen, with LAT vs. w/o LAT; with SEC vs. w/oSEC). A p-value <0.05 was considered as significant. Two multivariable models were built and stepwise backward logistic regression was performed to identify independent echocardiographic and clinical determinants for LAT and SEC. We included in the logistic regression model only those variables that at univariate analysis had a statistical difference with a p<0.1. Receiver operating characteristic (ROC) curves were built to identify cut-off values that could predict LAT and SEC. Statistical analyses were performedusing SPSS statistical software version19 (SPSS Inc., Chicago, Ill., USA).

A total of 403 patients are included in the present analysis.

Demographic and clinical data of the overall cohort are reported in [Table 1]. A longer than 3 weeks oral anticoagulation with either novel oral anticoagulants (NOAC) or Warfarin at therapeutic levels (INR>2) had been already performed in 30.0%.

A total of 262 patients (65%) had received no previous anticoagulation, 47 (11.7%) were treated with NOAC (rivaroxaban), 74 (18.4%) were anticoagulated with warfarin and had an INR>2 for >3 weeks at time of ECV (verified by either weekly laboratory testing or daily testing with a home-measurement-device), and 20 (5.0%) had received warfarin but at time of ECV had an INR<2 [Figure 1].

Figure 1. Flow diagram of patient’s baseline data and events, apportioned by anticoagulation management

---

---

AF: Atrial Fibrillation; OAC: oral anticoagulants; NOAC: novel oral anticoagulants; LVEF: Left Ventricular Ejection Fraction; LAT: Left Atrial Thrombosis; SEC: Spontaneous Echo-contrast; *Information available on 362 patients

[Table 1] reports clinical data according to the anticoagulation regimen. It appears that AF duration (paroxysmal and persistent) is unevenly distributed among the different categories of anticoagulated/non-anticoagulated patients. It emerges in fact, that a significantly higher rate of persistent AF (and significantly lower rate of paroxysmal AF) was present in anticoagulated patients (p=0.02). Furthermore, the subgroup of anticoagulated patients included a significantly higher rate of patients suffering from coronary artery disease (p=0.03). Moreover, patients receiving warfarin were significantly older (p<0.0001) and had a significantly higher CHA2DS2-VASc-score (p=0.001).

Clear evidence of LAT was present,at pre-ECV TEE, in 41 patients (10.2%) and SEC in 154 (38.2%).

A total of 27 patients (6.6%) had both LAT and SEC. Patients with evidence of LAT and/or SEC did not undergo ECV and were referred back to the treating cardiologist to optimize anticoagulation. Repeatedfollow-up imaging, starting after four weeks of optimized anticoagulation treatment, was recommended in all cases. All remaining patients (235/403; 58.3%) underwent at least one ECV attempt (data concerning sinus rhythm achievement are not reported). No acute thromboembolic events immediately after ECV were reported.

Patients under Warfarin and with INR> 2 had a trend for higher LAT rate (13.5%) and a significantly higher SEC rate (56.8%; p=0.002) [Table 1].

[Table 2]-[Table 3] show clinical data according to the presence of LAT/SEC.

At univariate analysis patients with LAT had a significantly lower left ventricular ejection fraction (LVEF%) (p=0.001). Patients with SEC were significantly older (p=0.04), had a significantly lower LVEF% (p<0.0001), higher CHADSVASC score (p<0.0001), and higher rate of CAD (p=0.03).

Two separate logistic regression models were built to identify determinants of LAT and SEC. LVEF% was the only echocardiographic independent predictor of LAT (p=0.02; OR=0.96; CI: 0.93-0.99; inverse relationship) and SEC (p<0.0001; OR:0.95; CI: 0.93-0.97; inverse relationship) [Table 4]. At ROC analysis a LVEF of 45% or lower showed a significant prediction of LAT and SEC (AUC: 0.7; p<0.0001; Sensitivity 75.0%- Specificity 60.0%). Moreover, therapeuticanticoagulation with warfarin (INR>2) (p=0.003; OR: 2.2; CI: 1.3-3.7) and CHADSVASC score (p<0.0001; OR=1.2; CI: 1.1-1.4) were both independent predictors of SEC. At ROC analysis aCHADSVASC score value of 3.5 was the cut-off to predict SEC (AUC: 0.7; p<0.0001; Sensitivity 64.0%- Specificity 60.0%).

Findings of our analysis can be summarized as follows:a) routine TEE prior to ECV shows clear signs of LAT/SEC in more than a third of referred AF patients independently by their anticoagulation regimen; b) the rate of LAT/SEC increases significantly with the decrement of LVEF% but does not seem to be independently related, at least in our analysis, to LA size; c) cumulative risk stratification methods, such as the CHADSVASC score, seem to have an independent predictive power particularly upon the risk of SEC; d) the risk of SEC increases for CHADSVASC scores above a 3.5 value; e) finally, even in anticoagulated patients the LAT and SEC rates remain higher than expected. In particular, therapeutic anticoagulation (INR>2) with Warfarin appears to be independently correlated with SEC occurrence.

A notable number of 262 patients (65%) had received no previous anticoagulation to hospital admission. Traceable reasons for no previous anticoagulation included a CHADSVASC score <3 for females and < 2 for males, non-compliance to anticoagulation, or absolute contraindication to anticoagulation resulting mainly from increased bleeding risk. Furthermore, a not negligible number presented to their outpatient doctor on the day of admission with a sudden unset of a previously untreated atrial fibrillation episode, which, due to the pronounced symptoms, immediately led to the patient being referred to our hospital for acute medication and subsequent TEE and ECV.Our tertiary care hospital mainly treats patients referred by their family doctor, or their outpatient cardiologist with a specific indication to ECV. While we are convinced that all the necessary efforts to anticoagulate the patient have been made properly, within a society dense with fragile and elderly patients it is quite common to encounter situations where permanent oral anticoagulation is contraindicated. We recommend continuation of oral anticoagulation for at least 4 weeks after successful ECV in all patients and discuss alternative options to permanent anticoagulation (e.g. left atrial appendage closure) with the further attending physician of all patients with a high stroke and high bleeding risk.

The study includes a heterogeneous group of patients referred from different facilities after being managed with different anticoagulation strategies. In this context, although the overall sample size is significant (in consideration of the fact that all patients had non-valvular AF and underwent pre-ECV TTE and TEE), pre-ECV treatment was not uniform and most of the patients had actually received no anticoagulation before being sent to our facility. Moreover, we do not have a structured follow-up on the referred patients and, as a matter of fact, we are not aware of the evolution of LAT and SEC in those patients that were denied ECV. Furthermore, the duration of AF did not emerge as a potential independent derteminant for SEC and LAT in our logistic regression model. The stronger impact of other variables such as LVEF, warfarinization, and CHADVASC score upon SEC/LAT may veil the real effect of AF duration. Some of those determinants, like LVEF, may actually be a surrogate of AF duration, suggesting there may exist a collinearity among the independent determinants. Unfortunately, there is no way to eliminate this collinearity from the present data.

Additionally, a stratification of patient groups into those with and those without anticoagulation for primary analyses could be usefull to gain additional information about independent risk factors but has not been performed to avoid a considerable decrease in power by reducing the sample size.

The occurrence of LAT and SEC is consistent, independently from the anticoagulation regimen adopted. Clinical and echocardiographic parameters can be used to predict the occurrence of LAT and SEC in patients with non-valvular AF. This information could be used in the future to select patients that, in spite of appropriate anticoagulation, should be evaluated with TEE before undergoing ECV.

Finally, although most of our present clinical decisions are based on our interpretation of TEE findings, in the very next future new methods of imaging analysis, based on artificial intelligence systems, will be made available for the daily clinical practice. Improved and automated image interpretation will allow us to better understand the LA and LAA anatomy and possibly detect specific anatomical features and early signs correlated to LAT and SEC occurrence [24].