Introduction

Atrial fibrillation (AF) is a common life-threatening arrhythmia affecting 1% to 2% of the general population. Thromboembolic complications constitute the most severe consequences of AF. They may be caused by a thrombus formation due to impaired blood flow in the left ventricle as a result of arrhythmia, most often in the left atrial appendage.1 So far, numerous clinical and echocardiographic predictors of left atrial appendage thrombus (LAAT) have been identified in patients on anticoagulant treatment. The CHADS2 and CHA2DS2-VASc scores are used for the assessment of thromboembolic risk in patients with nonvalvular AF. The CHA2DS2-VASc-RAF score additionally includes the type of AF and renal dysfunction as LAAT predictors, which appeared to improve the thromboembolic risk stratification.2-4 Despite optimal treatment with vitamin K antagonists and the currently preferred non–vitamin K oral antagonists (NOACs), LAAT still develops in 2% to 10% of patients on anticoagulant treatment.5-8 The gold standard for the exclusion of LAAT is transesophageal echocardiography (TEE).9,10 However, there are no unequivocal data on whether patients awaiting ablation or electric cardioversion should routinely undergo TEE, which is an invasive procedure that should be performed by trained and experienced personnel.11-14 Risk stratification according to sex is common in clinical practice. Sex-related differences in myocardial anatomy as well as the epidemiology, pathophysiology, and risk factors of heart diseases, including genetic cardiomyopathies, have been well described. Moreover, differences in the risk of stroke according to sex were reported in clinical trials.15-17 The aim of this study was to identify differences in the predictors of LAAT between men and women treated with NOACs (dabigatran or rivaroxaban).

Patients and methods

Study design and participants

In this retrospective study, we evaluated 1312 patients with AF from 3 centers in Poland, who underwent TEE before electrical cardioversion or catheter ablation between January 2013 and December 2019 and received continuous therapy with NOACs for at least 3 weeks. Patients taking apixaban (n = 56) were excluded from the study due to a small sample size. The final study population included 1256 individuals treated with dabigatran or rivaroxaban. The patients with moderate or severe mitral valve stenosis and those with a mechanical heart valve were excluded. The research protocol was approved by the ethics committee of each institution (17/2016). As this was an observational retrospective study, the patients’ written informed consent was not necessary.

Clinical, laboratory, and echocardiographic data were obtained retrospectively from medical records. The patients were divided into 3 groups (paroxysmal, persistent, or permanent AF) on the basis of a comprehensive analysis of the medical records. The patients were classified as having permanent AF after unsuccessful cardioversion during index hospitalization or if their original diagnosis of permanent AF was changed to long-standing persistent AF before electrical cardioversion or catheter ablation. These patients were classified as having permanent AF to distinguish them from those with persistent AF with a presumably lower AF burden. Vascular disease was defined as previous aortic plaque, myocardial infarction, or peripheral arterial disease. Kidney function was assessed using estimated glomerular filtration rate (eGFR) calculated with the Modification of Diet in Renal Disease Study formula.

Evaluation of thromboembolic risk

Thromboembolic risk in patients with AF was assessed using the following scores: CHADS2, CHA2DS2-VASc, R2CHADS2, and CHA2DS2-VASc-RAF. The CHA2DS2-VASc-RAF expands the risk assessment according to the CHA2DS2-VASc score by adding the type of AF and kidney function. The scoring systems used to estimate the risk of thromboembolic complications in patients with AF are described in detail in Table 1.

Table 1. Thromboembolic risk assessment scores in patients with atrial fibrillation

Parameter

Score

CHADS2 (max. 6 points)

R2CHADS2 (max. 8 points)

CHA2DS2-VASc (max. 9 points)

CHA2DS2-VASc-RAF (max. 25 points)

Risk factors

Heart failure

1

1

1

1

Hypertension

1

1

1

1

Diabetes mellitus

1

1

1

1

Vascular disease

1

1

Age 65–74 y

1

1

Stroke or transient ischemic attack

2

2

2

2

Age ≥75 y

1

1

2

2

Female sex

1

1

Creatinine clearance

eGFR <⁠60 ml/min/1.73 m2

2

eGFR <⁠56 ml/min/1.73 m2

2

Persistent AF

4

Permanent AF

10

Risk categories

Low

0

0

0 (men); 1 (women)

0–4 (men); 1–5 (women)

Medium

1

1

1 (men); 2 (women)

High

2

2

≥2 (men); ≥3 (women)

≥5 (men); ≥6 (women)

Abbreviations: AF, atrial fibrillation; CHADS2, congestive heart failure, hypertension, age over 75 years, diabetes mellitus, stroke or transient ischemic attack; CHA2DS2-VASc, congestive heart failure, hypertension, age over 75 years, diabetes mellitus, stroke or transient ischemic attack, vascular disease, age 65–74 years, sex category; CHA2DS2-VASc-RAF, congestive heart failure, hypertension, age over 75 years, diabetes mellitus, stroke or transient ischemic attack, vascular disease, age 65–74 years, sex category, renal dysfunction and AF type, eGFR, estimated glomerular filtration rate; R2CHADS2, creatinine clearance below 60 ml/min, congestive heart failure, hypertension, age over 75 years, diabetes mellitus, stroke or transient ischemic attack

Anticoagulant therapy

All patients received continuous therapy with dabigatran or rivaroxaban for at least 3 weeks before TEE, including the day TEE was performed. The drugs were administered according to the Summary of Product Characteristics.

Echocardiographic evaluation

All TEE examinations were performed by certified echocardiographers (second-degree accreditation in echocardiography of the Section of Echocardiography of the Polish Cardiac Society) within 48 hours before the scheduled procedure (electrical cardioversion or catheter ablation). The examinations were done using General Electric Vivid 7 or E95 Ultrasound System (General Electric, Milwaukee, Wisconsin, United States), EPIQ 7 Ultrasound Machine (Philips Medical Systems, Andover, Massachusetts, United States), or iE33 Ultrasound Machine (Philips Medical Systems) with an X72t TEE ultrasound transducer (Philips Medical Systems).

LAAT was defined as an independently mobile echo-dense structure that was distinct from the surrounding pectinate muscles or endocardium and that was detected in more than 1 imaging plane. Dense spontaneous echo contrast was defined as a dynamic “smoke-like” signal with a characteristic swirling motion, or a dynamic, gelatinous, precipitous echo density without a discrete mass, present throughout the cardiac cycle.

If LAAT was suspected, the images were evaluated by 2 echocardiographers to establish a unanimous diagnosis and enable a safe referral of the patient for electrical cardioversion or catheter ablation. Written informed consent for TEE was obtained from all patients. In cases with confirmed LAAT, a decision against the reversal of sinus rhythm was made. Decisions on further procedures in these patients were made on individual basis. The study end point was the presence of LAAT on TEE.

Statistical analysis

For qualitative variables, percentage values were calculated separately for each study group. The χ2 test was used to compare the groups. For quantitative variables, means with SDs were calculated. The quantitative variables with non-normal distribution were presented as median (interquartile range). Normally distributed variables were compared with the t test. Non-normally distributed variables were compared with the Mann–Whitney test.

A univariable and a multivariable logistic regression analysis were performed separately in women and men to evaluate associations between age, heart failure (HF), diabetes mellitus (DM), stroke / transient ischemic attack / peripheral artery disease, vascular disease, nonparoxysmal AF, creatinine clearance (eGFR <⁠60 ml/min/1.73 m2), and LAAT. A stepwise forward regression was used. Odds ratios (ORs) and 95% CIs were calculated. For all qualitative predictors, the absence of an event was considered as a reference point (0 value). In the logistic regression analysis, age was presented as quantitative data.

Two criteria for entering the variables into the model were applied. The Pearson correlation analysis was used for normally distributed variables and the Spearman correlation analysis was employed for variables with a distribution deviating from normal. The analyses investigated the relationship between the dependent variables and the predictors. The predictors whose correlation value with the dependent variable was below 0.05 were included in the model. Additionally, the model included the variables associated with the outcome known from the research literature.

To assess the 4 variables of CHADS2, CHA2DS2-VASc, CHA2DS2-VASc-RAF, and R2CHADS2 scores as well as to identify the optimal cutoffs for predicting LAAT, the receiver operating characteristic curve (ROC) analysis was used separately for women and men. The area under the curve (AUC) with 95% CI was the measure of each score. The optimal cutoff was the highest value of the Youden index (sensitivity + specificity – 1). For all calculations, a P value below 0.05 was considered significant. Statistical analysis was performed using the Statistica package (Tibco program, version 13.1, Warsaw, Poland).

Results

The study included 1256 patients (479 women [38.1%]) with AF hospitalized due to ablation and electrical cardioversion and treated with dabigatran or rivaroxaban. The mean (SD) age of the women was 60 (11.7) years, and of the men 67 (9.7) years. The clinical characteristics of the women and men are compared in Table 2.

Table 2. Baseline characteristics of all patients and separately of men and women

Variable

All patients (n = 1256)

Women (n = 479)

Men (n = 777)

P value

Age, y, mean (SD)

62.5 (11.4)

66.6 (11.7)

60.0 (9.7)

<⁠0.001a

Clinical characteristics, n (%)

Heart failure

263 (20.9)

86 (18.4)

175 (22.5)

0.08b

Hypertension

911 (72.5)

379 (79.1)

532 (68.5)

<⁠0.001b

Diabetes mellitus

231 (18.4)

91 (19)

140 (18)

0.66b

Stroke / TIA / embolism

93 (7.4)

41 (8.6)

52 (6.7)

0.14b

Vascular disease

34 (2.7)

13 (2.7)

21 (2.7)

0.99b

Type of AF

Paroxysmal

517 (41.2)

227 (47.4)

290 (37.3)

<⁠0.001b

Nonparoxysmal

739 (58.8)

252 (52.6)

487 (62.7)

Permanent

65 (5.2)

19 (4)

46 (5.9)

0.13b

Persistent

674 (53.7)

233 (18.6)

441 (56.8)

0.005b

Treatment, n (%)

Rivaroxaban

653 (52)

264 (55.1)

389 (50)

0.08b

Dabigatran

603 (48)

215 (44.9)

388 (49.9)

Reduced dose

110 (8.8)

63 (13.2)

47 (6)

<⁠0.001b

Thromboembolic risk

CHADS2, median (IQR)

1 (1–2)

1 (1–2)

1 (1–2)

0.002c

CHADS2 = 0, n (%)

248 (19.8)

80 (16.7)

168 (21.6)

0.03b

CHADS2 = 1, n (%)

535 (42.6)

199 (41.5)

336 (43.2)

CHADS2 ≥2, n (%)

473 (37.7)

200 (41.8)

273 (35.1)

CHA2DS2-VASc, mean (SD)

3.30 (1.56)

3.32 (1.56)

1.89 (1.49)

<⁠0.001a

CHA2DS2-VASc = 0, n (%)

135 (10.8)

0

135 (17.4)

<⁠0.001b

CHA2DS2-VASc = 1, n (%)

280 (22.3)

53 (11.1)

227 (29.2)

CHA2DS2-VASc ≥2, n (%)

841 (67)

426 (88.9)

415 (53.4)

CHA2DS2-VASc-RAF, median (IQR)

5 (3–8)

6 (3–9)

5 (2–7)

<⁠0.001c

R2CHADS2, median (IQR)

n = 1221; 2 (1–3)

n = 466; 2 (1–4)

n = 755; 1 (1–3)

<⁠0.001c

HASBLED, median (IQR)

1 (1–2)

2 (1-2)

1 (1–2)

<⁠0.001c

HASBLED <⁠3, n (%)

1105 (88)

391 (81.6)

714 (91.9)

<⁠0.001b

Laboratory data

Hemoglobin, g/dl, median (IQR)

n = 1229; 14.2 (13.2–15.2)

n = 460; 13.4 (12.6–14.2)

n = 769; 14.7 (13.8–15.5)

<⁠0.001c

Platelets, 109/l, median (IQR)

n = 1225; 217 (179–253)

n = 461; 230 (199–266)

n = 764; 208 (172–243)

<⁠0.001c

Creatinine, mg/dl, median (IQR)

1.04 (0.9–1.2)

0.97 (0.83–1.1)

1 (0.97–1.24)

<⁠0.001c

eGFR, ml/min/1.73 m2, median (IQR)

72 (58.2–90)

62.6 (52.4–90)

76.7 (63.4–90)

<⁠0.001c

eGFR <⁠60 ml/min/1.73 m2, n (%)

359 (28.6)

216 (54.9)

143 (18.4)

<⁠0.001b

Echocardiographic data

LA, mm, median (IQR)

n = 516; 45 (41–48)

n = 194; 43 (40–46)

n = 322; 46 (42–50)

<⁠0.001c

LVDD, mm, median (IQR)

n = 412; 51 (47–55.25)

n = 150; 48 (45–52)

n = 262; 46 (42–50)

<⁠0.001c

LVEF, %, median (IQR)

n = 616; 58 (50–60)

n = 241; 60 (55–60)

n = 375; 55 (45–55)

<⁠0.001c

LAAV, cm/s, median (IQR)

n = 769; 0.45 (0.3–0.7)

n = 302; 0.41 (0.28–0.59)

n = 467; 0.47 (0.3–0.7)

0.007c

End point

LAAT, n (%)

51 (4.1)

22 (4.6)

29 (3.7)

0.45b

a t test

b χ2 test

c Mann–Whitney test

Si conversion factors: to convert Hb to mmol/l multiply by 0.6206, creatinine to µmol/l muliply by 88.40.

Abbreviations: IQR, interquartile range; LA, left atrium; LAAT, left atrial appendage thrombus; LAAV, left atrial appendage flow velocity; LVDD, left ventricular diastolic diameter; LVEF, left ventricular ejection fraction; TIA, transient ischemic attack; others, see Table 1

LAAT was diagnosed in 51 patients (4.1%), including 22 women (4.6%) and 29 men (3.7%) (P = 0.45). The women with LAAT were older, were more likely to have HF and nonparoxysmal AF, received a reduced dose of anticoagulants, had a higher CHA2DS2-VASc-RAF score and lower left ventricular ejection fraction (LVEF) than the women without LAAT.

The men with LAAT were older, were more likely to have HF, nonparoxysmal AF, and DM, had higher CHADS2, CHA2DS2-VASc, R2CHADS2, and CHA2DS2-VASc-RAF scores, as well as lower eGFR and LVEF than the men without LAAT. A detailed comparison between the patients with and without LAAT is shown in Table 3.

Table 3. Baseline characteristics of the study groups according to sex and the presence of left atrial appendage thrombus

Variable

Women

P value

Men

P value

Without LAAT (n = 457)

With LAAT (n = 22)

Without LAAT (n = 748)

With LAAT (n = 29)

Age, y, mean (SD)

66.4 (9.7)

71.0 (8.9)

0.03a

59.7 (11.7)

65.9 (8.1)

0.002a

Clinical characteristics, n (%)

Heart failure

80 (17.5)

8 (36.4)

0.03b

156 (20.9)

19 (65.5)

<⁠0.001b

Hypertension

361 (79)

18 (81.8)

0.75b

510 (68.2)

22 (75.9)

0.38b

Diabetes mellitus

86 (18.8)

5 (22.7)

0.65b

127 (17)

13 (44.8)

<⁠0.001b

Stroke / TIA / peripheral embolism

39 (8.5)

2 (9.1)

0.93b

49 (6.6)

3 (10.3)

0.42b

Vascular disease

12 (2.6)

1 (4.6)

0.59b

20 (2.7)

1 (3.5)

0.80b

Type of AF, n (%)

Paroxysmal

225 (49.3)

2 (9.09)

<⁠0.001b

288 (38.5)

2 (6.9)

<⁠0.001b

Nonparoxysmal

232 (50.8)

20 (90.9)

460 (61.5)

27 (93.1)

Permanent

12 (3.6)

7 (31.8)

<⁠0.001b

38 (5.1)

8 (27.6)

<⁠0.001b

Persistent

220 (48.1)

13 (60)

0.32b

422 (56.4)

19 (65.5)

0.33b

Treatment, n (%)

Rivaroxaban

257 (56.2)

7 (31.8)

0.02b

375 (50.1)

14 (48.3)

0.84b

Dabigatran

200 (43.8)

15 (68.2)

373 (49.9)

15 (51.7)

Reduced dose

57 (12.5)

6 (27.3)

0.04b

46 (6.2)

1 (3.4)

0.55b

Thromboembolic risk

CHADS2, median (IQR)

1 (1–2)

2 (1–2)

0.18c

1 (1–2)

2 (1–3)

<⁠0.001c

CHADS2 =0, n (%)

78 (17.1)

2 (9.1)

0.40b

167 (22.3)

1 (3.4)

<⁠0.001b

CHADS2 =1, n (%)

191 (41.8)

8 (36.4)

329 (44)

7 (24.1)

CHADS2 ≥2, n (%)

188 (41.1)

12 (54.6)

252 (33.7)

21 (72.4)

CHA2DS2-VASc, median (IQR)

3 (2–4)

3.5 (3–5)

0.16c

2 (1–3)

3 (1–3)

<⁠0.001c

CHA2DS2-VASc = 0, n (%)

0

0

0.24b

134 (17.9)

1 (3.4)

0.005b

CHA2DS2-VASc=1, n (%)

53 (11.6)

0

223 (29.8)

4 (13.8)

CHA2DS2-VASc ≥2, n (%)

404 (88.4)

22 (100)

391 (52.3)

24 (82.8)

CHA2DS2-VASc-RAF, median (IQR)

6 (3–9)

10 (7–14)

<⁠0.001c

5 (2–7)

7 (7–13)

<⁠0.001c

R2CHADS2, median (IQR)

n = 444; 2 (1–3)

3 (2–4)

0.14c

n = 726; 1 (1–2)

3 (2–4)

<⁠0.001c

HASBLED, median (IQR)

2 (1–2)

2 (1–3)

0.11c

1 (1–2)

2 (1–2)

0.01c

HASBLED <⁠3, n (%)

375 (82.1)

16 (72.7)

0.27b

689 (92.1)

25 (86.2)

0.25b

Laboratory results

Hemoglobin, g/dl, median (IQR)

n = 439; 13.4 (12.6–14.2)

n = 21; 13.3 (12.6–14.3)

0.95c

n = 742; 14.7 (13.9–15.5)

n = 27; 14.3 (13.4–15.23)

0.08c

Platelets, 109/l, median (IQR)

n= 440; 230 (199–267)

n=21; 204 (195–258)

0.25c

n = 737; 208 (173–244)

n = 27; 172 (153–214)

0.008c

Creatinine, mg/dl, median (IQR)

0.96 (0.83–1.1)

1.05 (0.85–1.2)

0.47c

1.09 (0.97–1.23)

1.25 (0.99–1.36)

0.04c

eGFR, ml/min/1.73 m2, median (IQR)

62.6 (52.4–90)

55.10 (49.7–84.53)

0.44c

76.95 (63.98–90)

60.95 (56.8–90)

0.02c

eGFR <⁠60 ml/min/1.73 m2, n (%)

204 (44.6)

12 (54.5)

0.36b

130 (17.4)

13 (44.8)

<⁠0.001b

Echocardiographic data

LA, mm, median (IQR)

n = 184; 43 (40–46)

n = 10; 46.5 (45–47)

0.03c

n = 312; 46 (42–50)

n = 10; 48 (42–57)

0.26c

LVDD, mm, median (IQR)

n = 142; 48 (45–52)

n = 8; 49 (44.5–54)

0.80c

n = 252; 53 (49–57)

n = 10; 54 (51–58)

0.61c

LVEF, %, median (IQR)

n = 228; 60 (55–60)

n = 13; 55 (55–58)

0.01c

n = 360; 55 (47–60)

n = 15; 47 (40–55)

0.006c

LAAV, cm/s, median (IQR)

n = 287; 0.42 (0.29–0.58)

n = 15; 0.3 (0.15–0.73)

0.07c

n = 449; 0.48 (0.3–0.7)

n = 18; 0.36 (0.21–0.85)

0.32c

Abbreviations: see Tables 1 and 2

The results of the univariable regression analysis in men and women are presented in Table 4. In the multivariable logistic regression analysis, only nonparoxysmal AF was a predictor of LAAT in women (OR, 9.70; 95% CI, 2.24–41.97; P = 0.002). In men, the predictors of LAAT included HF (OR, 4.14; 95% CI, 1.82–9.40; P = 0.001), DM (OR, 2.64; 95% CI, 1.19–5.88; P = 0.002), nonparoxysmal AF (OR, 5.61; 95% CI, 1.29–24.46; P = 0.02), and eGFR <⁠60 ml/min/1.73 m2 (OR, 2.77; 95% CI, 1.25–6.13; P = 0.01). The results of the multivariable regression analysis in men and women are presented in Supplementary material, Tables S1 and S2.

Table 4. Univariable regression analysis in women and men of risk factors for left atrial appendage thrombus occurrence

Variable

Women

Men

OR

95% CI

P value

OR

95% CI

P value

Heart failure

2.69

1.09–6.63

0.03

7.21

3.29–15.82

<⁠0.001

Hypertension

1.20

0.40–3.62

0.75

1.47

0.62–3.48

0.38

Age

1.06

1.01–1.12

0.02

1.05

1.02–1.09

0.005

Diabetes mellitus

1.27

0.46–3.53

0.65

3.97

1.87–8.84

<⁠0.001

Stroke / TIA /peripheral embolism

1.17

0.26–5.20

0.84

1.80

0.53–6.18

0.35

Vascular disease

1.77

0.22–14.23

0.59

1.30

0.17–10.03

0.80

Nonparoxysmal AF

9.70

2.24–41.97

<⁠0.001

8.45

1.20–35.81

0.004

eGFR <⁠60 ml/min/1.73 m2

1.49

0.63–3.51

0.36

3.86

1.81–8.23

0.001

Reduced NOAC dose

2.63

0.989–7.001

0.05

0.55

0.07–4.1

0.56

Abbreviations: NOAC, non–vitamin K antagonist oral anticoagulants; OR, odds ratio; others, see Tables 1 and 2

Based on the receiver operating characteristic curve analysis, CHA2DS2-VASc-RAF had the highest diagnostic accuracy for predicting LAAT in women (Figure 1, Table 5). In men, all the scores had sufficient diagnostic power for predicting LAAT (Figure 2, Table 6).

Figure 1. Receiver operating curves for predicting left atrial appendage thrombus in women according to different thromboembolic risk scores

Abbreviations: see Table 1

Table 5. Receiver operating curve analysis for predicting left atrial appendage thrombus in women

Variable

AUC

95% CI

P value

Cutoff value

Youden index

True positives

False positives

False negatives

True negatives

Sensitivity

Specificity

CHADS2

0.580

0.460–0.700

0.19

2

0.134

12

188

10

269

0.545

0.589

CHA2DS2VASc

0.586

0.462–0.700

0.14

2

0.116

22

404

0

53

1.00

0.116

CHA2DS2VASc–RAF

0.775

0.675–0.875

<⁠0.001

10

0.394

13

90

9

367

0.591

0.803

R2CHADS2

0.592

0.476–0.708

0.12

2

0.144

17

279

5

165

0.773

0.372

Abbreviations: AUC, area under the curve; others, see Table 1

Figure 2. Receiver operating curves for predicting left atrial appendage thrombus in men according to different thromboembolic risk scores

Abbreviations: see Table 1

Table 6. Receiver operating curve analysis for predicting left atrial appendage thrombus in men

Variable

AUC

95% CI

P value

Cutoff value

Youden index

True positives

False positives

False negatives

True negatives

Sensitivity

Specificity

CHADS2

0.726

0.639–0.812

<⁠0.001

2

0.387

21

252

8

496

0.724

0.663

CHA2DS2VASc

0.734

0.644–0.823

<⁠0.001

3

0.441

21

212

8

536

0.724

0.717

CHA2DS2VASc–RAF

0.786

0.707–0.865

<⁠0.001

7

0.471

22

215

7

533

0.713

0.471

R2CHADS2

0.780

0.708–0.853

<⁠0.001

2

0.507

27

308

2

418

0.931

0.576

Abbreviations: see Tables 1 and 5

Discussion

Our study yields several major findings. First, women and men are at a similar risk of LAAT. Second, the frequency of LAAT in women and men is independent of the NOAC dose. Third, men and women differed in terms of the LAAT predictors and the predictive value of the thromboembolic risk scores.

The prevalence of LAAT was similar in women and men treated with dabigatran and rivaroxaban. Risk stratification according to sex is a common approach in clinical practice. There have been numerous studies assessing the risk of stroke depending on sex. Girijala et al15 analyzed sex-related differences in terms of the epidemiology of risk factors, symptoms, as well as methods and results of stroke treatment. Epidemiological data indicated a higher rate of stroke in older women, which raised the question of the neuroprotective role of sex hormones.16 In a retrospective study on the frequency of ischemic stroke in women and men with AF, Yoshida et al17 assessed sex-related differences in the risk factors for stroke as well as the anatomy and function of the left atrium. They concluded that for the same level of increase in the risk of stroke, worsening of left atrial function is greater in women than in men, which could explain a more frequent occurrence of ischemic stroke in women. In our study, women had a smaller left atrial size, smaller left ventricular diastolic diameter, and higher LVEF than men. Similarly, Shah et al18 observed that left atrial dilation, absence of severe mitral regurgitation, and lower LVEF are associated with an increased risk of LAAT.

HF is a component of all thromboembolic risk scores.19 In our study, the patients with LAAT significantly more often had HF. In the univariable logistic regression analysis, it was a strong predictor of LAAT both in men and women.

Di Mino et al20 conducted a meta-analysis including 72 studies and a total of 20 516 patients, in which they investigated the frequency of LAAT in individuals undergoing electric cardioversion or ablation. A higher frequency of LAAT was observed in women than in men (OR, 1.35; 95% CI, 1.04–1.75). The frequency of LAAT in women and men was independent of NOAC dose. Rivaroxaban was used by 52% of the patients, of which 55.1% were women. Dabigatran was used by 48% of the patients, of which 44.9% were women. However, the differences were not significant. A reduced dose of dabigatran and rivaroxaban was used by 13.2% of women and 6% of men (P <⁠0.001). The reduced dose was not a predictor of LAAT in either of the sexes. Similarly, Bertaglia et al7 studied a group of 414 patients treated with NOACs for more than 3 months, who underwent TEE before cardioversion or ablation. The reduced dose was not a predictor of LAAT either in men or in women.

Renal impairment is a documented risk factor for cardiovascular disease and it is associated with hypertension and DM, which are included in the CHADS2 and CHA2DS2-VASc scores. Arnson et al21 conducted a retrospective analysis of 85 116 patients with AF and concluded that advanced kidney disease was associated with a higher risk of stroke, death, and bleeding. Some studies2,3 identified renal dysfunction to be an important predictor of LAAT, enhancing the prognostic value of the CHA2DS-VASc score. In our study, eGFR below 60 ml/min/1.73 m2 was reported in 54.9% of the women and only 18.4% of the men. Impaired renal function was a predictor of LAAT only in men.

We found that the type of AF (ie, nonparoxysmal AF) was the strongest independent predictor of LAAT both in men and women. In men, additional independent predictors of LAAT were HF, DM, and impaired kidney function. In a retrospective study, Wei-dong et al22 assessed 705 of the 1346 patients with a low CHA2DS2-VASc score in whom TEE was performed before ablation or cardioversion. Nonparoxysmal AF was the strongest independent risk factor of LAAT / spontaneous echo contrast (OR, 3.766; 95% CI, 1.282–11.061; P = 0.02).

The guidelines of the European Society of Cardiology indicated that the type of AF does not affect the thromboembolic risk. However, some studies on this issue have been recently published. Dimitrova et al23 conducted a systematic overview of epidemiological studies on thromboembolic risk depending on the duration of AF, which were published in the MEDLINE and PubMed in the years 2005–2019. The authors emphasized a relatively low number of relevant studies, which may be due to methodological reasons (difficulty in a reliable assessment of the incidence of paroxysmal AF, which may be asymptomatic and thus underestimated or, on the other hand, it may recur frequently, resulting in a relatively long duration of arrhythmia). Importantly, in a meta-analysis of 12 studies including a total of 99 996 patients, Ganesen et al24 confirmed that the thromboembolic risk is higher in patients with nonparoxysmal than paroxysmal AF. Interesting conclusions were also reported in a retrospective study by Go et al.25 They assessed the thromboembolic risk in patients with paroxysmal AF depending on the total arrhythmia time. The patients with a longer total time of AF episodes had a higher risk of nonischemic stroke.

Study limitations

First, we could only indirectly identify the factors predisposing the patients with AF to thromboembolic complications by using LAAT as a surrogate end point. However, as the source of embolism in patients with AF is most often the LAA, the presence of LAAT can be considered a good surrogate for thromboembolic complications. A second limitation of the study is its restrictive nature. For example, detailed echocardiographic data (eg, LAAV, LVEF) were not available for all subjects, and hence, they could not be included in the multivariable analysis. Another limitation was the inability to determine the AF burden. In particular, the duration of arrhythmia in the patients was unknown, which did not allow us to distinguish between persistent and long-standing persistent AF. The small number of patients with LAAT, the inability to evaluate the frequency of indications for NOAC dose reduction, and the risk of noncompliance were further limitations of the study.

Conclusions

LAAT was present with a similar frequency in women and men despite using dabigatran and rivaroxaban. Nonparoxysmal AF was a strong predictor of LAAT in box sexes. Sex-related differences in the risk factors of LAAT and predictive value of thromboembolic risk scores were found. The predictors of LAAT in men were HF, DM, and eGFR below 60 ml/min/1.73 m2. In women, the highest predictive value of LAAT risk was revealed for CHA2DS2-VASc-RAF score, while in men, all the scores had a similar predictive value.