Transesophageal Echocardiographic Correlates of Clinical Risk of Thromboembolism in Nonvalvular Atrial Fibrillation

Study group

The design and the main results of the SPAF III study have been described in detail elsewhere (5,6). Patients with AF receiving aspirin therapy were prospectively stratified into risk groups using the scheme presented in Figure 1. High risk patients were characterized by one or more of the following: 1) women >75 years of age, 2) systolic hypertension >160 mm Hg measured on two occasions at entry, 3) impaired left ventricular function (clinical congestive heart failure within 100 days of entry or M-mode echocardiographic fractional shortening ≤25%), or 4) previous thromboembolism. Patients without these risk factors were further divided into those with a history of hypertension (moderate risk) and those without a history of hypertension (low risk) on the basis of a priori planned secondary analysis (6). Longitudinal follow-up of low and moderate risk patients with AF receiving aspirin, and high risk participants receiving aspirin plus low doses of warfarin (which proved ineffective in protecting against stroke) or adjusted-dose warfarin, confirmed prospectively the validity of this risk stratification scheme (4,5). Classification and treatment assignment were determined before the TEE examination and not influenced by its findings.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Observed thromboembolic rates in patients receiving aspirin according to risk strata among SPAF III participants undergoing TEE. Rate in high risk group is for patients receiving combination therapy. BP = blood pressure; CHF = congestive heart failure; FS = fractional shortening by precordial echocardiography.

Participation required signed informed consent according to federal and local regulations governing research involving humans.

TEE

Participants in the SPAF III study were encouraged to undergo TEE within 3 months of study entry, and 41% (n = 786) agreed to do so (48%, 42% and 37% of high, moderate and low risk participants, respectively). TEE studies were performed on average at 26 days after entry according to specific protocols for acquisition and interpretation (7). Commercially available ultrasound equipment was used, with biplane or multiplane transducers operating at a standard frequency of 5.0 MHz on gray-scale imaging. The TEE features were recorded according to standardized criteria among all 18 participating centers, and are described in detail elsewhere (7). Left atrial appendage (LAA) area was obtained by planimetry from the transverse (zero degree) view. Spontaneous echo contrast (SEC) was defined as a pattern of slowly swirling intracavitary echodensities imaged with gain settings adjusted to distinguish background noise and classified as dense when continuously present at standard gain. Aortic plaque was categorized as either simple (sessile) or complex based on features of mobility, ulceration, pedunculation or thickness ≥4 mm and could be located in the ascending, transverse or proximal descending segment (i.e., not just proximal to the origins of the cerebral vessels) (7). Analyses reported here are based on image interpretation at the local clinical laboratories. All TEE characteristics were concealed from the echocardiographers during recruitment and data acquisition, avoiding bias at the time of imaging interpretation.

About half of participants undergoing TEE were receiving warfarin before study entry (Table 1), and half of high risk patients were assigned to warfarin after enrollment. In order to characterize the frequency of LAA thrombus in the absence of warfarin effect (11), TEE studies carried out within 14 days of a patient receiving adjusted-dose warfarin (n = 311), either randomized or taking at baseline, were excluded from the analyses of LAA thrombus.

Although diabetes mellitus and intermittent (as opposed to continuous) AF were not found to be independent predictors of thromboembolism during aspirin therapy in SPAF I and SPAF II trials (8), these two factors have been associated with stroke risk in AF by other investigators (3,9). Thus, we took this opportunity, using SPAF III data, to reanalyze the role of diabetes and intermittent AF by correlating them with TEE features and thromboembolic risks.

Statistical analysis

Patient characteristics were compared across the three risk groups using a Mantel-Hantsel test for trend for categorical variables and analysis of variance for continuous variables. Individual groups were then compared using a chi-square test (or Fisher exact test if any expected cell value was <5) or two-tailed sample ttest. TEE characteristics independently predictive of risk group were determined by fitting a polychotomous logistic regression model with the moderate risk group as the reference group (EGRET statistical software, Cytel Corporation). All variables were placed in the model and terms that were statistically insignificant (log likelihood ratio test) were then eliminated (or set equal). A separate analysis was also done using a forward stepwise logistic regression model (log likelihood ratio test) to identify features of moderate and high risk patients (combined group) relative to low risk (SPSS/PC, SPSS Inc.). Significance was accepted at the p = 0.05 level and all tests were two-tailed. No adjustment was made for multiple comparisons.

Patients

Among the 786 participants undergoing TEE at entry, patient features differed substantially according to risk stratum (Table 1). As expected, high risk patients more frequently had the clinical and precordial echocardiographic features used as criteria for increased thromboembolic risk.

Correlates of thromboembolic risk

As indicated in Table 2, LAA thrombi increased in frequency from 3% to 8% to 15% (p < 0.001) between participants categorized as low, moderate and high risk, respectively. Dense SEC and peak LAA flow velocities were also univariately associated with risk stratum (p < 0.001), but the cross-sectional area of the LAA was not. The presence of aortic plaque, particularly plaque classified as complex, was more frequent in high risk patients with AF than those at low or moderate risk (Table 2). Figure 2illustrates the frequency of TEE findings versus thromboembolic risks. LAA abnormalities (presence of at least one of thrombus, dense SEC or reduced flow velocities, p < 0.001), and complex aortic plaque (p < 0.001) were associated with increased thromboembolic risk (Figure 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Frequency of TEE findings versus thromboembolic risk. Patients with left atrial abnormalities had at least one of thrombus, dense SEC or LAA flow velocities ≤20 cm/s. Gray bars= LAA abnormalities; solid bars= complex plaque.

Independent correlates of thromboembolic risk by multivariate analysis

As illustrated in Table 3, LAA thrombus (relative risk [RR] 2.7, p < 0.001) and reduced LAA peak flow velocity (RR 1.8, p = 0.003) were associated with moderate risk (history of hypertension without high risk features) relative to low risk status. Complex aortic plaque (RR 2.9, p < 0.001) and dense SEC (RR 2.7, p < 0.001) were more frequent in high risk than in moderate risk patients with AF. In patients with uncontrolled systolic hypertension as their only high risk feature, complex aortic plaque was significantly more frequent than in those with a history of hypertension (35% vs. 16%, p = 0.002). Independent TEE correlates of low versus moderate or high risk were LAA thrombus (RR 2.5, p = 0.04), dense SEC (RR 3.7, p < 0.001), LAA peak flow velocity ≤20 cm/s (RR 1.7, p = 0.008) and complex aortic plaque (RR 2.1, p < 0.001).

Characteristics associated with individual high risk features

In high risk patients, TEE features did not differ significantly among individual criteria conferring high risk status (Table 4). Figure 3illustrates the frequency of TEE findings grouped as LAA abnormalities (presence of at least one of thrombus, dense SEC or reduced flow velocities) and complex plaque in patients with a single high risk factor.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Frequency of TEE findings according to criteria classified as high risk. Participants with more than one high risk feature were excluded. Patients with left atrial abnormalities had at least one of thrombus, dense SEC or LAA flow velocities ≤20 cm/s. Gray bars= LAA abnormalities; solid bars= complex plaque.

Influence of diabetes, intermittence of AF and age among low risk participants

Low risk patients with and without diabetes had no significant differences in TEE features: the presence of LAA thrombi was 0% versus 3%, dense SEC was 0% versus 5% and complex aortic plaque was 17% versus 15%, between those with and without diabetes, respectively. Similarly, intermittent AF was not significantly associated with TEE features (Table 5). Among low risk patients with AF, age (assessed as a continuous variable) was associated with complex aortic plaque (72 vs. 65 years, p < 0.001) and to a lesser extent with dense SEC (72 vs. 66 years, p = 0.08). There were too few patients with LAA thrombus to assess its relation with age among low risk patients for meaningful comparison.

Study limitations

In multivariate models, the estimates of predictive strength of features with a low prevalence (e.g., LAA thrombi in low risk patients with AF) may be influenced by less specific but more prevalent variables (e.g., dense SEC). Hence, our estimates of the strength of correlations of LAA thrombi and dense SEC with thromboembolic risk may be affected by this interaction. Similarly, the low frequency of severe mitral regurgitation impaired the ability to accurately characterize its relation to thromboembolic risk. Even so, it seems clear that abnormalities of the left atrium predisposing to stasis or thrombus formation, or both, are important differences between patients with AF at low and those at high risk of thromboembolism.

Clinical implications

These TEE correlations strongly support a relation between left atrial stasis, thrombus formation and stroke risk in patients with nonvalvular AF and suggest that factors outside of the atrium may contribute to the thromboembolic risk, particularly in high risk patients with AF. It is possible that efforts to restore and maintain sinus rhythm in high risk patients with AF may not completely mitigate the increased risk of stroke if embolism from extracardiac sources is the cause.