The aim of this study was to investigate whether left atrial (LA) strain has incremental value over the CHA 2 DS 2 -VASc score for stratifying the risk for embolism in patients with atrial fibrillation (AF) and whether LA strain predicts poststroke mortality.
Consecutive patients with paroxysmal or persistent AF with acute embolism (82 patients) or without (204 controls) were prospectively enrolled. Global peak LA longitudinal strain during ventricular systole (LA S ) was assessed during AF rhythm. Global LA S was compared between the groups in the first cross-sectional study. Then, the 82 patients with acute embolism were prospectively followed during the second prospective cohort study.
Global LA S was lower in patients with acute embolism than in controls ( P < .001). Global LA S < 15.4% differentiated patients with acute embolism from controls, with an area under the curve of 0.83 ( P < .0001). In multivariate analysis, global LA S was independently associated with acute embolism (odds ratio, 0.74; 95% confidence interval, 0.67–0.82; P < .001) and had an incremental value over the CHA 2 DS 2 -VASc score ( P < .0001). Furthermore, 26 patients with acute embolisms died during a median follow-up period of 425 days. Global LA S independently predicted mortality after embolism.
In this observational study, LA strain provided incremental diagnostic information over that provided by the CHA 2 DS 2 -VASc score, suggesting that LA strain analysis could improve the current risk stratification of embolism in patients with AF. LA strain can also predict poststroke mortality.
Atrial fibrillation (AF) is associated with a substantial risk for stroke and systemic embolism. Embolic stroke is often fatal, and survivors are often left permanently disabled. Several novel oral anticoagulant drugs (OACs) have been developed to prevent stroke related to AF, and risk stratification for embolism and preventive strategies have become popular concerns. Various risk stratification schemes (RSSs) for embolisms have recently been developed. The CHA 2 DS 2 -VASc score is a risk factor–based RSS and is the most prevalent stratification scheme in clinical practice.
AF is associated with left atrial (LA) remodeling, including structural, functional, and electrical changes. Several studies have found that abnormal LA substrate or function is related to prior stroke. However, most risk factors that constitute the current RSSs are based on clinical characteristics, not on LA features.
Speckle-tracking echocardiography provides a precise assessment of the left atrium in patients with AF. Moreover, previous studies have shown that LA parameters are associated with cardiac morbidity and are robust predictors of cardiovascular outcomes in various populations. Therefore, we speculated that evaluation of LA function, including strain, would improve the predictive value of the current RSSs. Accordingly, in the present study we aimed to determine whether LA strain has incremental value for risk stratification of embolism over the CHA 2 DS 2 -VASc score in patients with AF. We also tested the hypothesis that LA strain predicts mortality after embolism.
We prospectively considered 93 patients with nonvalvular AF who were referred to one of three Japanese hospitals (Gunma University Hospital, Maebashi Red Cross Hospital, or the Geriatrics Research Institute and Hospital) between July 2011 and January 2013 for treatment of acute stroke or systemic embolism (acute embolism). The institutional medical ethics committees of the three participating hospitals approved the study protocol, and all patients provided written informed consent to participate.
All patients were assessed using echocardiography immediately after presentation. Neurologists or cardiologists carefully evaluated each patient for embolism; ischemic stroke was confirmed by a focal neurologic deficit of sudden onset, the presence of AF, and magnetic resonance imaging findings. Patients with transient ischemic attacks or negative imaging findings were excluded from the study. Systemic embolism was defined by computed tomography or angiography. Clinical stroke severity was assessed on the basis of Glasgow Coma Scale score <12 hours after admission.
We excluded 11 of the 93 patients because they had transient ischemic attacks (n = 1) or poor-quality echocardiographic images (n = 3). Patients with paroxysmal AF in sinus rhythm at the time of echocardiography (n = 7) were also excluded, because the reported LA strains measured in sinus rhythm seem different than during AF. The final population of 82 patients (stroke, n = 77; systemic embolism, n = 5) were classified as having either paroxysmal or persistent AF according to the guidelines of the American College of Cardiology, the American Heart Association, and the European Society of Cardiology. We also considered 332 consecutive control subjects with nonvalvular AF, but without acute embolisms, who presented to our echocardiography laboratory for routine evaluation during the same period. Of these, 129 were excluded because of cardiomyopathies (n = 3), decompensated heart failure (n = 4), sinus rhythm at the time of echocardiography (n = 104), or poor echocardiographic images (n = 18).
In the first cross-sectional study, LA function was compared between patients with acute embolism and controls. The patients were also compared with age-matched and gender-matched controls (1:1 ratio). Next, the 82 patients with acute embolism were prospectively followed.
Embolism Risk Score
Embolism risk was assessed using the CHA 2 DS 2 -VASc score, which assigns one point for congestive heart failure within the past 3 months or left ventricular (LV) ejection fraction < 40%; one point for hypertension; two points for age > 75 years; one point for diabetes; two points for stroke, transient ischemic attack, or systemic embolism; one point for vascular disease (prior myocardial infarction and peripheral arterial disease); one point for age 65 to 74 years; and one point for female sex. The two points derived from the acute embolism event causing the present admission were not added to the patient risk scores during the evaluation; risk scores (including the CHA 2 DS 2 -VASc score) were calculated on the basis of the events leading to the embolic event precipitating the current admission.
Each patient was placed in the left lateral decubitus position and assessed using echocardiography with either the Vivid 7 or Vivid i ultrasound system (GE Medical Systems, Milwaukee, WI), equipped with 2.5-MHz transducers. LV volume and LA volume (LAV) were quantified according to the recommendations of the American Society of Echocardiography. LAVs were calculated using the biapical area-length method and subsequently indexed to body surface area (LAV index [LAVI]). The LA emptying fraction was defined as [(LAV max − LAV min )/LAV max ] × 100. Transmitral flow (E wave and deceleration time) and mitral annular tissue Doppler (E′ and S′) velocities were measured. The Doppler value recorded was the mean of three beats.
Three consecutive cardiac cycles were recorded and averaged, and the frame rate was set to 60 to 80 frames/sec. The analysis was performed offline using customized software (EchoPAC PC; GE Medical Systems). The LA endocardial border was manually traced in both four-chamber and two-chamber views. After manual adjustment of a region of interest covering the full thickness of the myocardium, the software divided the left atrium into six segments and automatically scored the segmental tracking quality. The software rejected segments with inadequate image quality and excluded them from analysis. Longitudinal strain curves were generated for each of the 12 LA segments in the four-chamber and two-chamber views. Because two segments of the LA roof demonstrated lower longitudinal strain curves than those of the other four, they were excluded from both the four-chamber and two-chamber views. Global peak LA longitudinal strain during ventricular systole (LA S ) was then measured by averaging the values obtained in the eight other LA segments. An experienced echocardiologist, who was unaware of the patients’ information, analyzed all echocardiographic values. Among the 6,840 echocardiographically evaluated LA segmental strain values, adequate tracking was possible in 6,710 (98.1%).
Patients with acute embolism were followed from the day of admission to death or last contact; the follow-up rate was 98% as of December 2013. The primary outcome was all-cause mortality. The secondary outcome was the modified Rankin scale (mRS) score at discharge, which is widely used to evaluate the functional outcomes of patients with stroke. The interobserver reliability of the mRS was excellent, at κ = 0.82.
Reproducibility of LA Strain
The reproducibility of LA strain was assessed in 10 randomly selected patients. Intraobserver and interobserver agreement was evaluated after the same observer and another experienced reader repeated the analysis 4 weeks later. The interclass correlation coefficient values for intraobserver and interobserver LA strain variability were 0.97 and 0.95, respectively.
All continuous variables are presented as mean ± SD, unless otherwise specified. Comparisons between groups were performed using χ 2 , Kruskal-Wallis, or Mann-Whitney U tests. Normality was evaluated using the Shapiro-Wilk W test. Agreement among the mRS findings was assessed using κ statistics. Baseline group differences were adjusted using general linear models. We constructed receiver operating characteristic (ROC) curves to determine the diagnostic ability of global LA S to identify acute embolism. The cutoff was obtained using the Youden index. The area under the curve for each parameter was compared using paired analyses. Associations between LA parameters and acute embolisms were assessed using univariate and multivariate logistic regression analyses. Conditional logistic regression analysis was used for matched populations. The prespecified variables of anticoagulant drugs, CHA 2 DS 2 -VASc score, age, sex, LAVI, LA emptying fraction, and global LA S were included in the multivariate models regardless of their associations in the univariate analysis, on the basis of our own hypothesis and previous findings. The incremental value was assessed by comparing the global χ 2 value for each model. We assessed the effect of global LA S on end points using Kaplan-Meier analysis, and the prognostic value of global LA S was determined using the Cox proportional-hazards model. The proportional-hazards assumption, assessed using the goodness-of-fit approach for each independent variable, was not violated. A two-sided P value < .05 was accepted as indicating statistical significance. All data were analyzed using SPSS version 21.0 (SPSS, Chicago, IL) and MedCalc version 220.127.116.11 (MedCalc Software, Mariakerke, Belgium).
Table 1 shows the clinical characteristics of the patients. Those with acute embolism were older, more frequently women, and less likely to be treated with OACs than controls; other medications were similar between the groups, including antiplatelet drugs. CHA 2 DS 2 -VASc scores before the current events were higher in patients with acute embolism than in controls.
|Clinical Characteristic||Controls (n = 203)||Patients with acute embolism (n = 82)||P|
|Age, y||72 ± 10||79 ± 10||<.001|
|Women||56 (28%)||40 (49%)||.001|
|Body mass index, kg/m 2||22.7 ± 3.4||21.8 ± 3.0||.029|
|Persistent AF||172 (86%)||70 (86%)||1.000|
|Systolic blood pressure, mm Hg||123 ± 20||134 ± 22||.001|
|Heart rate, beats/min||80 ± 20||87 ± 21||.026|
|Recent congestive heart failure within the past 3 mo||30 (15%)||4 (5%)||.025|
|LV ejection fraction < 40%||24 (12%)||8 (10%)||.683|
|Hypertension||126 (62%)||55 (67%)||.497|
|Age ≥ 75 y||85 (42%)||58 (71%)||<.001|
|Age 65–75 y||70 (35%)||17 (21%)||.023|
|Diabetes mellitus||40 (20%)||21 (26%)||.269|
|Prior stroke/embolism||46 (23%)||20 (24%)||.758|
|Peripheral artery disease||5 (3%)||4 (5%)||.285|
|Old myocardial infarction||13 (6%)||4 (5%)||.785|
|CHA 2 DS 2 -VASc score before current events||3.0 ± 1.6||3.8 ± 1.5||<.001|
|Glasgow Coma Scale score on admission||—||12.7 ± 2.7||—|
|eGFR, mL/min/1.73 m 2||63 ± 25||64 ± 22||.839|
|Brain natriuretic peptide, pg/mL||191 (113–300)||233 (147–338)||.103|
|Anticoagulants||147 (72%)||22 (27%)||<.001|
|Antiplatelets||49 (24%)||24 (29%)||.454|
|β-blockers||59 (29%)||15 (18%)||.073|
|ACE inhibitors/ARBs||92 (46%)||37 (45%)||1.000|
|Loop diuretics||82 (41%)||25 (31%)||.137|
|Statins||42 (21%)||16 (20%)||.872|
The median duration between admission and echocardiographic evaluation of patients with acute embolisms was 1 day (interquartile range, 0.0–2.0 days). The septum was slightly, but significantly, thickened in these patients, whereas LV mass index did not differ between the groups ( Table 2 ). Because more women had acute embolism, LV end-diastolic volumes were smaller in the embolic patient group, but the ejection fractions were similar between the groups. E′ velocity was significantly lower and transmitral E-wave velocity was similar, with increased E/E′ ratios, in these patients compared with controls, suggestive of higher filling pressures. Neither the maximum LAVI (LAVI max ) nor the minimum LAVI (LAVI min ) between the groups, whereas the LA emptying fraction and global LA S were lower in the embolic patients than in the controls. These results suggest that patients with acute embolism had impaired LA reservoir function compared with controls. The LA emptying fraction and global LA S remained significantly different between the groups after adjusting for age and sex ( P < .001 for both). Comparisons of the clinical characteristics between the patients with acute embolism and age-matched and gender-matched controls are presented in Supplemental Tables S1 and S2 (available at www.onlinejase.com ).
(n = 203)
|Patients with acute embolism |
(n = 82)
|Septal wall thickness, mm||10 ± 2||11 ± 2||.032|
|LV mass index, g/m 2||111 ± 29||111 ± 33||.809|
|LV end-diastolic volume, mL||72 ± 28||56 ± 28||<.001|
|LV ejection fraction, %||54 ± 11||55 ± 12||.606|
|E wave, cm/sec||91 ± 21||94 ± 25||.482|
|Deceleration time, msec||171 ± 51||157 ± 70||.002|
|E′, cm/sec||6.6 ± 1.9||5.9 ± 1.6||.002|
|S′, cm/sec||5.0 ± 1.3||4.5 ± 1.2||.002|
|E/E′ ratio||15.2 ± 6.6||17.3 ± 7.1||.007|
|LAVI max , mL/m 2||60 ± 21||58 ± 22||.391|
|LAVI min , mL/m 2||44 ± 19||48 ± 22||.203|
|LA emptying fraction, %||28 ± 13||20 ± 11||<.001|
|Global LA S , %||18.9 ± 6.0||12.6 ± 3.7||<.001|
Differentiation of Patients with Acute Embolism from AF Controls
CHA 2 DS 2 -VASc scores and global LA S distinguished patients with acute embolism from controls, with areas under the curve of 0.64 ( P < .0001) and 0.83 ( P < .0001) ( P < .0001 vs CHA 2 DS 2 -VASc score; Figure 1 and Supplemental Figure S1 [available at www.onlinejase.com ]). Global LA S < 15.4% (95% confidence interval, 14.7–15.8) identified patients with acute embolism with 83% sensitivity and 75% specificity (positive and negative likelihood, 3.30 and 0.23, respectively).
Independent and Incremental LA Strain Values for Risk Stratification of Acute Embolism Compared with CHA 2 DS 2 -VASc Score
In the analysis involving unmatched controls, univariate analysis identified OAC drugs, CHA 2 DS 2 -VASc score, age, female sex, LA emptying fraction, and global LA S as significant contributors to acute embolism, but not LAVI max or LAVI min ( Table 3 ). Global LA S remained an independent discriminator in each multivariate model. The odds ratios for LAVIs were <1 in the multivariate models, reflecting the higher frequency of recent congestive heart failure in controls than in patients with acute embolism in this population (15% vs 5%, P = .025). Furthermore, the LA emptying fraction and global LA S provided incremental diagnostic value over CHA 2 DS 2 -VASc score ( Figure 2 ). These results suggest that assessing LA function provided additional predictive information about acute embolism.
|Variable||Univariate||Multivariate 1||Multivariate 2||Multivariate 3|
|OR (95% CI)||P||OR (95% CI)||P||OR (95% CI)||P||OR (95% CI)||P|
|Anticoagulant use||0.14 (0.08–0.25)||<.001||0.13 (0.06–0.28)||<.001||0.12 (0.06–0.26)||<.001||0.12 (0.06–0.25)||<.001|
|CHA 2 DS 2 -VASc score||1.35 (1.14–1.61)||<.001||0.90 (0.65–1.25)||.537||0.87 (0.63–1.19)||.379||0.91 (0.66–1.25)||.563|
|Age||1.08 (1.05–1.12)||<.001||1.08 (1.03–1.13)||.002||1.07 (1.02–1.13)||.004||1.06 (1.01–1.11)||.015|
|Female sex||2.50 (1.47–4.25)||.001||0.96 (0.41–2.26)||.924||1.07 (0.46–2.53)||.870||0.84 (0.36–1.94)||.677|
|LAVI max||1.00 (0.98–1.01)||.468||0.97 (0.95–0.99)||.003||—||—||—||—|
|LAVI min||1.01 (1.00–1.02)||.189||—||—||0.97 (0.95–0.99)||.013||—||—|
|LA emptying fraction||0.95 (0.92–0.97)||<.001||—||—||—||—||0.98 (0.95–1.01)||.204|
|Global LA S||0.75 (0.70–0.81)||<.001||0.71 (0.64–0.79)||<.001||0.71 (0.64–0.78)||<.001||0.74 (0.67–0.82)||<.001|
Similarly, among the matched controls ( Supplemental Table S3 ; available at www.onlinejase.com ), OAC drugs, LA emptying fraction, and global LA S were significantly associated with embolism in univariate analysis. OAC drugs and global LA S remained independent discriminators throughout the three multivariate analyses.
Ten of the patients with acute embolism (12.2%) underwent emergency recanalization (mechanical thrombectomy or pharmacologic thrombolysis) after admission. The mRS scores ( Figure 3 ) were higher among patients with global LA S below the median value of 12.0% than among those with values above the median (4.0 ± 1.8 vs 2.8 ± 1.9, P = .003), suggesting worse functional outcomes at discharge. There was a weak (correlation coefficient = 0.229) but significant positive correlation between global LA S and Glasgow Coma Scale score ( P = .038). In patients with acute embolism, higher LA strain was associated with better levels of consciousness. The frequency of emergency recanalization did not differ between the groups (9.8% vs 14.6%, P = .737). The difference in mRS score remained significant after adjusting for age and sex ( P = .023).
There were 26 all-cause deaths among patients with acute embolism (31.7%) over a median follow-up period of 425 days (interquartile range, 136–603 days), and specific causes of death are shown in Table 4 . A Kaplan-Meier analysis showed that survival was reduced among patients with global LA S below the median value of 12.0% (log-rank P = .007; Figure 4 ). In the univariate Cox proportional-hazard analyses, increased mortality was related to age (hazard ratio, 1.08; P = .003), lower Glasgow Coma Scale score (hazard ratio, 0.80; P < .001), and lower global LA S (hazard ratio, 0.85 per 1%; P = .014). Neither sex nor anticoagulant use predicted death after acute embolism. Because the number of deaths was limited, we tested the robustness of global LA S in progressive models, as previously described ; global LA S remained an independent predictor in each bivariate prognostic model ( Table 5 ).