Demographics
According to the presence of atrial fibrillation related stroke, the 373 patients were divided into stroke (169 patients) and non-stroke (204 patients) group. Of the 169 patients with stroke, 109 (64.5%) were males and 60 (35.5%) were females. Of the 204 patients without stroke, 121 (59.3%) were males and 83 (40.7%) were females. 60 (15.7%) patients were with low CHA2DS2-VASc score (Table 1).
Compared with the patients without stroke, patients with stroke were older (median age: 66 (57-70) vs 62 (54-69), P=0.006), higher in CHA2DS2-VASC score (median score: 3 (2-4) vs 2 (1-3), P =0.002) and HASBLED score (median score: 3 (2-3) vs 2 (1-3), P < 0.001), a higher rate of hypertension history (66.3% VS 51.0%, P=0.003), diabetes history (22.5% VS 14.2%, P=0.038) and heart failure (49.1% VS 21.1%, P<0.001). No statistical differences were observed in gender, types of atrial fibrillation, alcohol use, smoking history, coronary heart disease history and peripheral vascular disease history between the two groups. (Table 1)
Comparison for Patients with Low CHA2DS2-VASc Score
To investigate the predictability of factors found in the entire cohort, we perform sensitivity analysis only using patients with low CHA2DS2-VASC score.
60 patients in this study were with low CHA2DS2-VASc score, 21 (35.0%) in the stroke group and 39 (65.0%) in the non-stroke group. 11 males were in the stroke group (52.4%) and 26 males were in the non-stroke group (66.7%). The mean age of stroke group was 54.5±6.5 years, and that of non-stroke group was 52.7±10.2 years.
The demographics between the two groups showed statistically significant differences in history of hypertension (3 (14.3%) vs 0 (0%), P=0.039), heart failure (3 (14.9%) vs 0 (0%), P=0.039), and HASBLED score (2 (1-2) vs 1 (0-1), P< 0.001). (Table 3)
Direct bilirubin (5.2μmmol/L (3.9-8.0) vs 3.8μmmol/L (2.6-4.6), P=0.013), indirect bilirubin (8.9μmmol/L (6.4-12.9) vs 5.9μmmol/L (4.0-8.4), P=0.008), and anion gap (14.3 mmol/L (11.2-15.7) vs 11.9 mmol/L (10.4-13.8), P=0.013) were higher for stroke group than that for non-stroke group.(Table 4)
In the examination of transesophageal and transthoracic echocardiography, only the LVEF value (59.8% (55.7%-66.4 %) vs 63.9% (61.3%-67.7%), P=0.031) between the two groups was statistically significant. (Table 4).
Laboratory Parameters
Compared with non-stroke group, stroke group had a larger mean platelet volume (9.3 fL (8.5-10.4) vs 8.9 fL (8.1-10.0), P=0.010), higher γ Glutamyl transferase (26 U/L (17-41) vs 20 U/L (16-29), P<0.001) and anion gap (13.8±2.5 vs 12.7±2.5 mmol/L, P<0.001). Stroke patients were lower in thyrotropin releasing hormone (TSH) (1.78 mIU/L (1.12-2.94) vs 2.25 mIU/L (1.36-3.40), P=0.021) (Table 2)
Transthoracic and Transesophageal Ultrasound
Aortic annulus diameter in stroke group was smaller than non-stroke group (31 mm (28-33) vs 32 mm (29-34), P=0.008). Patients with stroke were higher in early mitral valve forward flow velocity (EMV) (82 cm/s (67-101) vs 77 cm/s (62-92), P=0.030) and left atrial diameter (38 mm (36-43) vs 37 mm (33-41), P=0.008). They were also with a larger diameter measured by 45° (18 mm (16-20) vs 17 mm (15-19), P=0.018) and 135° (19 mm (17-22) vs 18 mm (16-22), P=0.012) for left atrial appendage opening. Left atrial appendage emptying velocity (LAAEV) is lower in stroke group (43 cm/s (28-61) vs 52 cm/s (33-75), P=0.002) (Table 2).
Multivariate Analyses for the Predictors of Cardioembolic Stroke Due to Atrial Fibrillation
CHA2DAS2 VASC score is based on heart failure, hypertension, diabetes, peripheral vascular disease, gender and age. To avoid the collinearity between predictors, CHA2DAS2 VASC score was selected into multiple binary logistic regression instead of using heart failure, hypertension, diabetes, peripheral vascular disease, gender and age.
Due to the correlation between 45°, 90° and 135° diameters of the left atrial appendage, it is decided to only use the 135° diameter in the multiple logistic regression. Similarly, only direct bilirubin was included in the multiple logistic regression if direct bilirubin and indirect bilirubin are both P<0.01 in the univariate analysis.
Except for CHA2DAS2 VASC score, other factors with P < 0.1 in univariate analysis were include in the multiple logistic regression. The final model included CHA2DS2-VSC score, mean platelet volume,γGlutamyl transpeptidase, anion gap, alanine aminotransferase, thyroid-stimulating hormone, aortic diameter, e-peak deceleration time(EDT), EMV, left atrial diameter, 135° left atrial appendage opening measured diameter, left atrial appendage emptying rate.
In multiple logistic regression, CHA2DAS2 VASC score (OR 1.22 (95%CI 1.04-1.43), P=0.016), anion gap (OR 1.19 (95%CI 1.08-1.30), P < 0.001), EDT (OR 1.01 (95%CI 1.00-1.01), P=0.001) and the LAAEV (OR 0.99 (95%CI 0.97-0.99), P=0.013) were risk factors for predicting stroke in patients with atrial fibrillation. (Table 5)
For patients with low CHA2DAS2 VASC score, direct bilirubin, anion gap, EDT, LV Ejection fraction and EMV were included in the multiple binary logistic regression. The results showed that anion gap (OR 1.35 (95%CI 1.03-1.77), P=0.028) and EDT (OR 1.01 (95%CI 1.00-1.02), P=0.043) were associated with stroke. (Table 6)
ROC Analysis for Predictors of Cardioembolic Stroke Due to Atrial Fibrillation
To investigate the predictability of the statistically significant factors in multivariate logistic regression, we compared two models using AUC of ROC curve.
Model 1: Predictor included CHA2DS2-VASc score
Model 2: Predictors included anion gap, e-peak deceleration time, left atrial appendage emptying rate and CHA2DS2-VASc score in patients with atrial fibrillation
ROC curve showed that model 2 had a higher AUC for predicting stroke onset (AUC 0.70 (95%CI 0.64-0.75)) compared to model 1 (AUC 0.59 (95%CI 0.54-0.65)) (Figure 1 (a)).
Considering the effects in patients with low CHA2DAS2 VASC score, we did a sensitivity analysis using the same selection criteria.
Model 3: Predictor included low CHA2DS2-VASc score
Model 4: Predictors included anion gap, e-peak deceleration time and CHA2DS2-VASc score in low CHA2DS2-VASc score patients with atrial fibrillation
For patients with low CHA2DAS2 VASC score, ROC curve analysis also showed that model 4 (AUC 0.74 (95% CI 0.61-0.88)) had a higher AUC compared to model 3 (AUC 0.55 (95% CI 0.40-0.70)) (Figure 1 (b)).