Left atrial (LA) strain as a marker for discrimination of risk for stroke and transient ischemic attack (TIA) in patients with atrial fibrillation and low-risk CHADS 2 scores (≤1) has yet to be examined.
Patients with atrial fibrillation, stroke or TIA, and CHADS 2 scores ≤ 1 before their events were identified retrospectively from a large single-center stroke registry and compared with age-matched and gender-matched controls. Antihypertensive use and echocardiographic parameters including chamber volumes and left ventricular mass and LA peak negative and positive strain and strain rate were compared between groups.
Fifty-seven patients meeting entry criteria were identified. Patients demonstrated significantly lower left ventricular ejection fractions, larger LA dimensions, and larger LA volume indexes (24.4 ± 11.9 vs 32.3 ± 13.3 mL/m 2 , P = .012) compared with controls. Both peak negative LA strain (−3.2 ± 1.2% vs −6.9 ± 4.2%, P < .001) and peak positive LA strain (14 ± 11% vs 25 ± 12%, P < .001) were significantly reduced in patients compared with controls. Peak negative LA strain was significantly associated with stroke by binary logistic regression (odds ratio, 2.15; P < .001).
In patients with low-risk CHADS 2 scores, atrial fibrillation, and stroke or TIA, reduced LA strain is a potentially sensitive maker for increased risk for stroke or TIA. These results suggest that LA strain may have potential as a tool for helping guide the decision for or against oral anticoagulation in this group of patients.
Since its emergence as an effective prophylaxis against cardioembolic stroke in patients with atrial fibrillation (AF), oral anticoagulation has endured as a therapeutic triumph. Warfarin decreases annual stroke rates for patients with nonrheumatic AF from 4.5% to 1.4% compared with placebo, for an overall stroke reduction of 60%. Risk for thromboembolism, however, is not evenly distributed among all patients with AF. In the absence of concomitant risk factors, the incidence of stroke in patients with AF is similar to that in the general population. As the sum of stroke risk factors increases, stroke frequency increases proportionately.
This heterogeneity of risk has led to the promulgation of numerous strategies for risk stratification, in which established stroke risk factors (heart failure, hypertension, older age, diabetes, and previous stroke or transient ischemic attack [TIA]) are added together to produce a total point score. Although risk increases with higher scores, stroke rates in lower risk patients vary up to fourfold depending on which scoring system is applied. Perhaps as a way to resolve the uncertainty generated by such variable results, current guidelines recommend that anticoagulation or antiplatelet therapy be individualized for patients with AF and only one other stroke risk factor. Although severe left ventricular (LV) dysfunction is recognized as a significant adverse predictor, no scheme at present incorporates any other marker of cardiac structure or function. Although elevated left atrial (LA) dimension is a common finding in AF and is associated with increased risk for ischemic stroke, its utility as an independent predictor of thromboembolic stroke risk in the setting of AF remains controversial.
We hypothesized that the addition of a quantitative functional marker of thromboembolic risk to current classification schemes might increase their robustness as a guideline for anticoagulation and in particular improve decision making for low-risk patients. We further hypothesized that LA strain would be a strong candidate for such a marker. We used a widely accepted classification system, the CHADS 2 score, to identify a cohort of patients with AF at relatively low risk for stoke or TIA. We compared echocardiographic dimensions, systolic function, and LA strain in these patients with these values in matched controls to determine whether that measure or any other echocardiographic variable would produce a significant difference between the two groups.
Patients and Controls
The Hartford Hospital stroke registry records a standard series of clinical and demographic variables for all patients presenting to the neurology service for evaluation for stroke or TIA. Since 2003 (the first year digitized echocardiograms were permanently stored), the registry has recorded a total of 6,190 individual patient encounters; of these, the diagnosis of stroke or TIA was confirmed by clinical assessment of a board-certified neurologist combined with computed tomographic or magnetic resonance imaging findings in 5,490 patients; stroke or TIA was ruled out in 1,163 patients. A history of paroxysmal AF or electrocardiographic evidence of AF within 24 hours of the event was confirmed in 615 patients. Within this group, 75 patients with imaging or clinical findings suggestive of thromboembolic stroke or TIA, CHADS 2 scores ≤ 1 before the index events, and technically adequate transthoracic echocardiograms within 24 hours of presentation were identified retrospectively. Five patients whose echocardiograms were technically inadequate for strain analysis were excluded. Because the following conditions independently qualify patients for oral anticoagulation or produce LA dysfunction, patients with LV ejection fractions < 35% ( n = 8), mitral stenosis ( n = 4), or the presence of a prosthetic heart valve or valve repair ( n = 6) were excluded, leaving 57 patients available for analysis. Controls were selected from patients within the registry ( n = 48) or the inpatient echocardiography laboratory database ( n = 9) with electrocardiographically documented AF or documented histories of paroxysmal AF with otherwise identical inclusion criteria but without evidence or history of stroke or TIA or any history of oral anticoagulation before the index echocardiographic assessment. Patients and controls were matched for sex, age (±3 years), and aspirin use before presentation. Patients underwent standard, digitized echocardiography within 72 hours of presentation. The study was approved by the Hartford Hospital Institutional Review Board.
Age, sex, international normalized ratio (INR), and CHADS 2 and CHA 2 DS 2 -VASc scores immediately before stroke or TIA were recorded. CHADS 2 scores were calculated by adding 1 point for congestive heart failure, hypertension, age ≥ 75 years, and diabetes mellitus, and CHA 2 DS 2 -VASc scores were compiled by adding 1 more point for female sex, documented vascular disease, or age 64 to 75 years. Body surface area was recorded for each patient and control. All echocardiographic images were obtained from Siemens Sequoia 512 (Siemens Medical Solutions USA, Inc., Mountain View, CA), Philips iE33 (Philips Medical Systems, Andover, MA), or Philips 7500 (Philips Medical Systems) sonographs, translated into Digital Imaging and Communications in Medicine format, and transferred to an offline analysis system (Heartlab; Agfa, Hackensack, NJ). Echocardiographic dimensions were measured, and LV mass index was calculated. LV ejection fraction was measured from apical four-chamber and two-chamber views using Simpson’s biplane method. LA volume was calculated using the biplane method of discs. Maximum LA volume was measured just before mitral valve opening, minimum LA volume was measured just before mitral valve closure, and LA emptying volume was calculated as the difference. All echocardiographic measurements were made by a level 3–trained, board-certified echocardiographer blinded to the presence or absence of stroke or TIA. Strain measurements were made by a single blinded investigator (V.M.R.).
LA strain was measured from Digital Imaging and Communications in Medicine images imported into a previously validated grayscale analysis system (Axius Velocity Vector Imaging; Siemens Medical Solutions USA, Inc) ( Figure 1 ). Velocity Vector Imaging is a software program that translates any Digital Imaging and Communications in Medicine echocardiographic image into a grayscale image in a manner analogous to speckle tracking. Angle-independent vectors are constructed along the entire length of the atrial (or ventricular) wall (solid arrow in Figure 1 ). Vectors representing both relative velocity and direction throughout the cardiac cycle are displayed as a sequence of tracking arrows. The program displays segmental or global velocity, strain, or strain rate ( Figure 1 ). LA strain has been evaluated in both sinus rhythm and AF, and LA strain in AF measured by both methods is closely correlated.
The endocardial border of the left atrium was traced in the apical four-chamber and two-chamber views. Manual readjustments were made as necessary to ensure that endocardial borders were accurately traced. Regional LA strain and strain rate data were obtained from the annular, mid, and superior segments along the septal lateral, anterior, and inferior LA walls using apical four-chamber and two-chamber views ( Figure 1 ). Global LA strain data were computed by taking the average of longitudinal strain data obtained from a total of 12 LA segments (six segments from each view). Peak positive LA strain and strain rate were measured during LV systole (dotted arrow in Figure 1 ), and peak negative strain and strain rate were measured at LV end-diastole or during LA contraction (solid arrow in Figure 1 ). Mean frame rates for each echocardiogram were identical for both patients and controls (43 Hz) and were similar to those used in previous studies. Assessment of LA strain was regarded as suboptimal ( n = 5) when either strain analysis could not be performed for at least four of the six LA segments in each view or a theoretically unacceptable value or values were obtained.
All data are expressed as mean ± SD or as percentages. Using previous published data describing the distribution of LA strain and strain rate for patients with AF and stoke, and accepting a 20% decrease in peak systolic strain as significant, we estimated that a sample of 51 patient-control pairs would provide 80% power with an α error of 0.05. After testing each distribution for normality, differences between relevant demographic, clinical, and echocardiographic variables were compared between patients and controls using paired t tests. Differences in the frequency of categorical variables were tested using χ 2 tests. Correlation was tested using Pearson’s correlation coefficient. Binary logistic regression was constructed using the presence of stroke or TIA as the dependent variable and all significant echocardiographic variables as covariates in the model. Statistical calculations were performed using SPSS version 19.0 (SPSS, Inc., Chicago, IL).
The mean age of the entire cohort was 65 years ( Table 1 ); 30 patient-control pairs (53%) were men. TIA was documented in 10% ( n = 6) of patients; five patient-control pairs (9%) were classified with CHADS 2 scores of 0. In the remaining 52 patient-control pairs, hypertension was ubiquitous and was the factor responsible for producing a CHADS 2 score of 1 in all but one patient and two controls ( Table 1 ). Mean CHA 2 DS 2 -VASc scores were higher (2.0 ± 0.9 for both patients and controls) in large part because women constituted nearly half the study cohort. Seventy-five percent of pairs (43 of 57) were matched by both CHADS 2 and CHA 2 DS 2 -VASc scores, and 30% of patients had CHA 2 DS 2 -VASc scores ≤ 1. A minority of patient-control pairs (34%) were taking aspirin. Eleven patients (19%) had histories of warfarin use before stroke or TIA; in four patients, warfarin had been discontinued before a surgical procedure, the INR was ≤1 in each patient, and the stroke occurred perioperatively. The remaining patients all presented with subtherapeutic INRs documented at the time of stroke or TIA (mean, 1.54 ± 0.28; range, 1.06–1.81). Patients demonstrated significantly lower LV ejection fractions, larger LA dimensions, and larger LA volume indexes (LAVIs) compared with controls ( Table 1 ). There was no significant difference in the frequency of moderate or severe mitral regurgitation or in the frequency of paroxysmal AF between groups ( Table 1 ).
|Variable||No stroke/TIA |
( n = 57)
( n = 57)
|Age (years)||65 ± 7||65 ± 7||.34|
|BSA (m 2 )||2.0 ± 0.25||2.0 ± 0.30||.17|
|INR||1.08 ± 0.27||1.14 ± 0.25||.21|
|Hypertension||50 (89%)||49 (86%)||.98|
|Diabetes mellitus||1 (2%)||2 (3.5%)||.56|
|Antihypertensive use||48(84%)||40 (70%)||.07|
|LVEDD (cm)||4.3 ± 0.8||4.5 ± 0.7||.14|
|LVESD (cm)||2.7 ± 0.8||3.0 ± 0.8||.13|
|LA dimension (cm)||3.7 ± 0.6||4.1 ± 0.7||.008|
|LAVI (mL/m 2 )||24.4 ± 11.9||32.3 ± 13.3||.012|
|LV mass index (g/m 2 )||94 ± 34||106 ± 44||.107|
|LVEF (%)||65 ± 7||61 ± 9||.03|
|LA emptying fraction||43 ± 25||30 ± 43||.038|
|Moderate or greater MR||5.3%||8.8%||.54|
|Peak negative LA strain (%)||−6.9 ± 4.2||−3.2 ± 1.2||<.001|
|Peak negative LA strain rate (sec −1 )||−1.9 ± 0.7||−1.2 ± 0.6||<.001|
|Peak positive LA strain (%)||25 ± 12||14 ± 11||<.001|
|Peak positive LA strain rate (sect −1 )||1.8 ± 0.7||1.2 ± 0.6||<.001|
Peak negative LA strain and LA strain rate were significantly lower in controls compared with patients, and peak positive strain and strain rate were significantly higher ( Table 1 ). Differences between patients and controls were similar for patients with permanent AF and paroxysmal AF ( Table 2 ). All four strain indexes were weakly but significantly correlated across the entire cohort, with patients generally skewed toward the upper right end of the scatter graph (lower peak negative LA strain and higher LAVI; Table 3 , Figure 2 ).
|Peak negative LA strain (%)||−2.9 ± 2.1||−6.4 ± 2.9||<.001|
|Peak negative LA strain rate (sec −1 )||−0.07 ± 0.57||−1.6 ± 1.2||.001|
|Peak positive LA strain (%)||8.2 ± 5.5||19 ± 12||<.001|
|Peak positive LA strain rate (sec −1 )||0.8 ± 0.5||1.5 ± 0.6||.001|
|Peak negative strain (%)||−3.4 ± 1.5||−6.8 ± 5.4||.001|
|Peak negative strain rate (sec −1 )||−1.4 ± 0.53||−1.9 ± 0.78||.012|
|Peak positive strain (%)||20 ± 12||29 ± 12||.003|
|Peak positive strain rate (sec −1 )||1.5 ± 0.6||2.0 ± 0.7||<.001|
|Strain Index||Correlation with LAVI ( r )||P|
|Peak negative LA strain||0.24||.012|
|Peak negative LA strain rate||0.38||<.001|
|Peak positive LA strain||−0.30||.002|
|Peak positive LA strain rate||−0.35||<.002|