The value of diastolic function parameters in the prediction of left atrial appendage thrombus in patients with nonvalvular atrial fibrillation

A retrospective cohort study design was implemented. We queried the echocardiography laboratory database at Rush University Medical Center to identify all consecutive adult patients with nonvalvular AF who underwent a TEE to “rule-out” left atrial appendage thrombus between January 1, 2005 and December 31, 2009. Of those, we only included patients who had a previous transthoracic echocardiogram (TTE) within 1 year of the TEE. Patients with atrial flutter, without intervening episodes of atrial fibrillation, were not included. We excluded patients with valvular AF due to mitral stenosis and those with conditions known to alter E and e’ velocities; namely mitral regurgitation greater than 2+ in severity (on a 0 to 4 scale), post mitral valve surgical or percutaneous intervention, and post orthotopic heart transplantation status. Patients with isolated aortic valvular disease, aortic valve prostheses and right-sided valvular heart disease were not excluded [1].

An expert board-certified (NBE) echocardiographer (RD), who was blinded to the TTE and clinical data, reviewed all TEE images to determine the presence or absence of LAAT [LAAT(+) and LAAT(−)], SEC and depressed left atrial appendage emptying velocity (<40 cm/sec) by pulsed-wave Doppler. LAAT was defined as a circumscribed and uniformly echodense intracavitary mass distinct from the underlying left atrial or left atrial appendage endocardium and the pectinate muscles, and present in more than one imaging plane [9]. Left atrial appendage sludge, defined as a dynamic gelatinous, precipitous echodensity, without a discrete mass, present throughout the cardiac cycle was categorized as LAAT [10, 11]. SEC was defined as dynamic “smoke-like” echoes with the characteristic swirling motion with optimal gain setting during the entire cardiac cycle [12].

All TTEs were reviewed offline by a board-certified (NBE) echocardiographer who was blinded to TEE and clinical data (HG). The mitral inflow early-diastolic pulsed-wave Doppler velocity (E) and the lateral mitral annulus tissue Doppler early-diastolic velocity (e’) were measured (cm/sec) and the E:e’ ratio was calculated (Figure 1). We did not analyze e’ velocities sampled from the medial mitral annulus, since this was not routinely obtained according to our laboratory protocol at the time when the study TTEs were performed (2005–2009). The Doppler measurements were obtained by averaging data from 3–5 consecutive beats. The left ventricular septal and posterior wall thicknesses and end-diastolic as well as end-systolic internal dimensions were measured from 2-dimentional and M-mode TTE images [13]. The left ventricular mass was calculated using the Devereux formula [14]. The left ventricular systolic and diastolic volumes and LVEF were measured using the biplane Simpson’s method or the Teichholz formula only when the former was not feasible due to suboptimal apical views [13, 15]. The left atrial dimensions (antero-posterior, medio-lateral and supero-inferior) were measured and the left atrial volume was calculated using the formula [4/3π (D1/2) (D2/2) (D3/2)] where each “D” represents one of three atrial dimensions [13]. All volume and mass measurements were indexed to the body surface area. The heart rhythm during TTE acquisition was determined from examination of the rhythm strip on the screen display.

Finally, a detailed chart review was performed to collect data on demographics, comorbidities, anticoagulant and antiplatelet use, and INR (international normalized ratio). Congestive heart failure (CHF) was defined as any clinical heart failure or LVEF < 40%. Date-of-onset and chronicity of AF were determined from patients’ health records. Paroxysmal AF was defined as AF spontaneously reverting to sinus rhythm within 7 days from onset. Persistent AF was defined as AF lasting more than 7 days, but less than 6 months, or any AF events terminated by electrical or chemical cardioversion or radiofrequency ablation within 6 months from onset. Permanent AF was defined as lasting more than 6 months [1]. The CHADS₂ score was calculated from the sum of risk predictors of CHF, hypertension, age ≥75, diabetes mellitus, stroke or transient ischemic attack (TIA); weighing each by “1” except for prior stroke/transient ischemic attack (TIA) which was weighed by “2” [16]. The CHA₂DS₂-VASc score was calculated from the sum of the risk factors of CHF, hypertension, age 65–74 or ≥75 years, diabetes mellitus, stroke/TIA, vascular disease, female gender; weighing each by “1” except for stroke/TIA and age ≥75 which were weighed by “2” [17]. The primary outcome of the study was TEE-identified LAAT. The secondary outcome was TEE-identified SEC.

The two-tailed Student’s t-test was used to compare normally-distributed continuous variables which were expressed as mean ± standard deviation (SD). The Mann–Whitney-U test was used to compare continuous variables that did not adhere to a normal distribution which were expressed as median (interquartile range). The Shapiro-Wilk test was used to confirm the normality of data. The chi-square (χ²) test was used to compare categorical variables which were expressed as frequency [n (%)]. The Spearman’s method was used to evaluate linear correlations.

Multivariate logistic regression analysis models were used to determine clinical and echocardiographic predictors of outcome measures independent of known confounders. Risk was expressed as odd-ratios (OR) with 95% confidence intervals (CI). The Hosmer-Lemeshow test was used to test regression models’ goodness of fit.

The receiver operating characteristic (ROC) methodology was used to analyze the discriminatory capacity of various predictors of LAAT. ROC analyses were expressed as curve plots with the associated area under the curve (AUC) with 95% CI and a P value representing the likelihood of the null hypothesis (AUC = 0.5). A two-tailed P value ≤0.05 was considered statistically significant in all analyses. PASW-18 software (SPSS, Inc. - Chicago, IL) was used for all data analyses with the exception of the comparisons between ROC curves for which STATA-11 (College Station, TX) was used. The study was funded by an internal research grant and was approved by the Rush University Medical Center institutional review board.

The mean E:e’ among LAAT(+) patients was significantly higher than those who were LAAT(−) [16.6 ± 6.1 vs. 12.0 ± 5.4, respectively; P = 0.001]. Conversely, the e’ velocity was significantly lower among LAAT(+) subjects [6.5 ± 2.1 vs. 9.1 ± 3.2, P = 0.006] (Figure 2, Table 2). Among LAAT(+) subjects, none (0%) had a normal E:e’ of ≤8; whereas 16 (84%) had an E:e’ ≥ 12, indicative of elevated LVFP [18]. Furthermore, there was a stepwise increase in the prevalence of LAAT with increasing E:e’ and decreasing e’ velocity (Figure 3). Additionally, LAAT(+) subjects had a significantly higher LAVI and left ventricular volume index but lower LVEF (Table 2). Not surprisingly, there was a modest, but highly statistically significant, linear correlation between the CHADS₂ score and diastolic indices of the E:e’ and e’ velocity, with Spearman’s correlation coefficients “r” of 0.34 (P < 0.001) and −0.32 (P < 0.001), respectively. Since only a minority of the patients (33%) were in sinus rhythm at the time of TTE, and thus had no A wave in the mitral inflow Doppler waveforms, it was not feasible to classify the patients based on their grade of diastolic dysfunction (typically graded from I to IV).

A multivariate logistic regression analysis demonstrated that the CHADS₂ score is associated with LAAT independent of warfarin use [OR = 1.47 per one point CHADS₂ score increment (CI = 1.04-2.1), P = 0.03], while warfarin use had a borderline association with LAAT (Table 3; Model-1). When E:e’ was added to Model-1, it was independently associated with LAAT [OR = 1.14 per 1 point increment (CI = 1.05-1.2), P =0.002] and negated the effect of the CHADS₂ score (Model-2, Table 3). To ensure that this association is not simply a confounder to impaired left ventricular function and left atrial volume, we added LVEF and LAVI to the model forming Model-3, in which E:e’, LVEF and LAVI were independently associated with LAAT, whereas the value of the CHADS₂ and warfarin use was further negated (Table 3, Model-3).

Similarly, when e’ velocity then LVEF and LAVI were sequentially introduced to Model-1 (Models 4 and 5), e’ velocity was independently “protective” of LAAT after adjusting for other covariates [OR = 0.68 for each 1 cm/sec increment in e’ velocity, P = 0.007] while the CHADS₂ score was not (Table 3; Model-5). The Hosmer and Lemeshow test demonstrated a good fit of Models 3 and 5 (Table 3). Nearly identical results were obtained when analyzing the CHA₂DS₂-VASc score instead of CHADS₂ score.

Additionally, E:e’ and e’ velocity were also associated with LAAT independent of the individual components of CHADS₂ score [E:e’ odds-ratio = 1.07 per 1 point increment (CI = 1.01-1.14), P = 0.02; e’ velocity odds-ratio = 0.87 per 1 cm/sec increment (CI = 0.77-0.98), P = 0.03]. These models maintained a good model fit (Hosmer and Lemeshow test P values = 0.30 and 0.23, respectively). Similar findings were noted with the CHA₂DS₂-VASc score.

One-hundred fourteen subjects (38%) had significant SEC by TEE, including all 19 subjects (100%) with LAAT (Table 2). These patients, as compared to those without SEC, had significantly higher mean E:e’ [14.2 (±5.6) vs. 11.4 (±5.4), P = 0.001] and lower e’ velocity [7.7 (±2.5) vs. 9.6 (±3.4) cm/sec, P < 0.001]. In multivariate logistic regression analyses, E:e’ and e’ velocity were associated with SEC independent of LVEF, LAVI, CHADS₂ score, and warfarin therapy (Table 4), with similar results when substituting the CHA₂DS₂-VASc for the CHADS₂ score.

The receiver operator characteristics (ROC) curves demonstrated that the E:e’ and e’ velocity have good discriminatory capacity in predicting LAAT with respective areas under the curve (AUC) of 0.72 and 0.74, which trended to be larger than the 0.65 AUC associated with the CHADS₂ score (P = 0.17 and 0.052, respectively), as shown in Figure 4. In this population, the ROC curve point-coordinates identified an E:e’ value of ≥9.4 to have 100% sensitivity and 38% specificity for LAAT; whereas E:e’ ≥ 15 was associated with a specificity of 78% at the expense of a low sensitivity (32%). An e’ velocity ≤10 cm/sec was associated with 100% sensitivity for LAAT (Figure 4). Only a CHADS₂ score of zero was “protective” from LAAT.