Prognostic role of left atrial strain and its combination index with transmitral E-wave velocity in patients with atrial fibrillation

Left atrial (LA) strain can reflect LA remodeling and is reduced in atrial fibrillation (AF) patients with prior stroke. This study sought to examine the ability of LA strain in predicting subsequent stroke event in AF and also evaluated whether E/LA strain could predict cardiovascular (CV) events in these patients. In 190 persistent AF patients, we performed comprehensive echocardiography with assessment of LA strain. There were 69 CV events including 19 CV death, 32 hospitalizations for heart failure, 3 myocardial infarctions, and 15 strokes during an average follow-up of 29 months. Multivariate analysis showed old age, chronic heart failure, increased left ventricular (LV) mass index, and increased E/LA strain were associated with CV events and decreased LA strain was associated with subsequent stroke event. The addition of E/LA strain and LA strain to a model containing CHA2DS2-VASc score and LV function significantly improved the values in predicting CV events and subsequent stroke event, respectively. In conclusion, E/LA strain and LA strain were respectively useful in predicting CV events and subsequent stroke event in AF. E/LA strain and LA strain could provide incremental values for CV outcome and subsequent stroke outcome prediction over conventional clinical and echocardiographic parameters in AF, respectively.

Scientific RepoRts | 6:17318 | DOI: 10.1038/srep17318 transmitral E-wave velocity (E) to left ventricular diastolic parameters can effectively predict CV outcomes [14][15][16][17] . LA strain has been demonstrated to be associated with left ventricular diastolic function 18 . We hypothesize the combination index, E/LA strain, is also a useful parameter in predicting CV events in AF patients. Hence, this study was designed to investigate whether E/LA strain and LA strain could respectively predict adverse CV events and subsequent stroke event in patients with AF.

Methods
Study patients. This observational cohort study prospectively and consecutively included patients with persistent AF referred for echocardiographic examinations at Kaohsiung Municipal Hsiao-Kang Hospital from April 2010 to June 2012. Persistent AF was defined as AF lasting for at least 7 days. Patients with moderate and severe mitral stenosis (n = 5), moderate and severe aortic stenosis or regurgitation (n = 4), severe mitral regurgitation (n = 5), and inadequate echocardiographic visualization (n = 11) were excluded. Additionally, four patients who had no beat fulfilling the requirements of index beat in the stored cardiac cycles were also excluded. Finally, 190 AF patients were included in this study. The study protocol was approved by our Institutional Review Board and all enrolled patients gave written, informed consent.
Echocardiographic evaluation. The echocardiographic examination was performed by one experienced cardiologist with a VIVID 7 (General Electric Medical Systems, Horten, Norway). The cardiologist was blinded to the clinical data. Pulsed tissue Doppler imaging was obtained with the sample volume placed at the lateral and septal corners of the mitral annulus from the apical 4-chamber view. Early diastolic mitral annulus velocity (E') was averaged from septal and lateral ones. The wall filter settings were adjusted to exclude high-frequency signals and the gain was minimized. Left ventricular ejection fraction (LVEF) was measured using the modified Simpson's method. Left atrial volume was measured using the biplane area-length method 19 . Left atrial volume index (LAVI) was calculated by dividing left atrial volume by body surface area.
LA strain measurement. The endocardial border was manually defined using a point-and-click technique. An epicardial surface tracing was automatically generated by the system, creating a region of interest, which was manually adjusted to cover the full thickness of left atrium. Time-strain plot was produced automatically by the software. Global LA strain during the reservoir phase was estimated by taking the average of longitudinal strain data obtained from the apical four-chamber and two chamber projections 10,13,20 . Data from a total of 12 LA segments (annular, mid, and superior segments along the septal, lateral, anterior, and inferior LA walls using apical four-chamber and two-chamber images) were averaged to determine global longitudinal LA strain at the end of left ventricular ejection (LA reservoir phase). Assessment of LA strain was accepted if at least four of the six LA segments in each view could be measured clearly. We used cine loops to determine which beat would be calculated. The raw ultrasonic data, including 15 consecutive beats from the apical four-chamber and two-chamber views, was recorded and analyzed offline using EchoPAC software (EchoPAC version 08; GE-Vingmed Ultrasound AS GE Medical Systems).
Left ventricular dimensions, LVEF, LAVI, and LA strain were measured from the index beat method 21,22 . Because their measurements were easy and rapid, the E, E-wave deceleration time, and E' were obtained from five beats 23 and then averaged for later analysis. If the cardiac cycle length was too short to complete the diastolic process, this beat was skipped. Thus, the selection of E, E-wave deceleration time, and E' was not always consecutive. In addition, heart rate was determined from five consecutive beats.
Collection of demographic, medical, and laboratory data. Demographic and medical data including age, gender and history of diabetes mellitus, hypertension, coronary artery disease, stroke, and chronic heart failure were obtained from medical records or interviews with patients. Laboratory data including total cholesterol and triglyceride were also collected. In addition, information regarding patient medications during the study period was obtained from medical records.
Definition of CV events. CV events were defined as CV mortality, hospitalization for heart failure, myocardial infarction, and stroke. Hospitalization for heart failure was defined as admission due to dyspnea with chest radiographic evidence of pulmonary congestion and treatment with intravenous diuretics. CV events were ascertained and adjudicated by two cardiologists with disagreement resolved by adjudication from a third cardiologist from the hospital course and medical record. If patients had multiple CV events, only the first event was coded. However, if patients died after episodes of heart failure, myocardial infarction, or stroke during the same admission, they were coded as CV death. In patients reaching the study end points, they were followed until the first episode of adverse events. The other patients were followed until March 2014. Statistical analysis. SPSS 18.0 software (SPSS, Chicago, IL, USA) was used for statistical analysis. Data were expressed as mean ± standard deviation, percentage, or median (25 th -75 th percentile) for follow-up period. Continuous and categorical variables among groups were compared by one-way analysis of variance (ANOVA) followed by a post hoc test adjusted with a Bonferroni correction and Chi-square test, Scientific RepoRts | 6:17318 | DOI: 10.1038/srep17318 respectively. The relationship between two continuous variables was assessed by a bivariate correlation method (Pearson's correlation). The significant variables in the univariate analysis were selected for multivariate analysis. Time to the adverse events and covariates of risk factors were modeled using a Cox proportional hazards model with forward selection. A significant improvement in model prediction was based on the − 2 log likelihood ratio statistic, which followed a difference in − 2 log likelihood value and the P value was based on the incremental value compared with the previous model. Kaplan-Meier survival plots were calculated from baseline to time of adverse events and compared using the log-rank test. Stepwise multiple linear regression analysis was employed to identify the determinants of LA strain and E/LA strain. All tests were 2-sided and the level of significance was established as P < 0.05. Table 1 shows the comparison of clinical and echocardiographic characteristics according to the tertile of LA strain. There were significant differences in the age, prevalence of coronary artery disease and chronic heart failure, CHA 2 DS 2 -VASc score, heart rate, anticoagulant use, LAVI, left ventricular end-systolic and end-diastolic dimensions, LVEF, E' , E/E' , and LA strain among patients with different tertiles.

Results
The follow-up period to CV events was 29 (25 th -75 th percentile: 18-36) months in all patients. Sixty-nine CV events were documented during the follow-up period, including CV death (n = 19), hospitalization for heart failure (n = 32), myocardial infarction (n = 3), and stroke (n = 15). A Cox proportional hazards regression analysis for CV events is shown in Table 2. In the multivariate analysis, old age, the presence of chronic heart failure, increased LVMI, and increased E/LA strain (hazard ratio [HR], 1.182; 95% confidence interval [CI], 1.086 to 1.286; P < 0.001) were independently associated with increased CV events.
The follow-up period to subsequent stroke event was 30 (25 th -75 th percentile: 23-41) months in all patients. Eighteen subsequent stroke events were documented during the follow-up period. A Cox proportional hazards regression analysis for subsequent stroke event is shown in Table 3. In the multivariate analysis, only decreased LA strain (HR, 0.844; 95% CI, 0.745 to 0.955; P = 0.007) was independently associated with increased subsequent stroke event.
To find the appropriate cut-off values to E/LA strain and LA strain as predictors of the outcomes, we created several models using different cut-off values of E/LA strain and LA strain. Using the − 2 log likelihood value to select the model with the best performance, the model using E/LA strain > 6.00 m and LA strain < 16.50% had the best performance in predicting the adverse events. Figure 1 illustrates the Kaplan-Meier curves for CV event-free survival and subsequent stroke-free survival in study patients.
The incremental values of E/LA strain and LA strain in outcome prediction are shown in Table 4. The addition of E/LA strain and LA strain to a Cox model containing CHA 2 DS 2 -VASc score, LAVI, LVEF, and E/E' significantly improved the values in predicting CV events and subsequent stroke event respectively. However, the addition of LA strain and E/LA strain to a Cox model containing CHA 2 DS 2 -VASc score, LAVI, LVEF, and E/E' did not improve the values in predicting CV events and subsequent stroke event, respectively.

Discussion
In the present study, we evaluated the association of E/LA strain with CV events and LA strain with subsequent stroke event in patients with AF. We found that increased E/LA strain and decreased LA strain were independently associated with an increase in CV events and subsequent stroke event in AF, respectively. The E/LA strain and LA strain could respectively add significant incremental values beyond the conventional clinical and echocardiographic parameters in prediction of CV events and subsequent stroke event.
Diastolic dysfunction increases left ventricular filling pressure and is the primary mechanism responsible for the clinical findings of heart failure. Echocardiography is the most valuable tool for the noninvasive evaluation of diastolic function 24 . Several approaches on the basis of Doppler modalities have been proposed as useful methods for the evaluation of left ventricular diastolic function 25,26 . By combining these parameters with those obtained from the mitral flow curve, a more precise estimation of left ventricular filling pressure has been achieved 27,28 . E/E' was reported to be related well to mean pulmonary capillary wedge pressure 28 and the ratio of E to global diastolic strain rate had also been proposed as a marker of elevated left ventricular filling pressure 29 . Ersbøll M et al. showed the ratio of E to global early diastolic strain rate (E'sr) was independently associated with an adverse outcome in patients with myocardial infarction 14 . Our previous study also showed that E/E'sr was a useful parameter in predicting adverse cardiac events in AF 17 . LA strain has demonstrated to be associated with left ventricular diastolic function 18 . Hence, E/LA strain should be a good parameter in the evaluation of left ventricular filling pressure. In fact, the present study showed increased E/LA strain was highly correlated with increased E/E' , a good parameter of left ventricular filling pressure 28 , in the multivariate analysis. The LA strain has been reported to be useful in predicting CV prognosis in AF patients with acute embolism 30 . In the present study, although decreased LA strain and increased E/LA strain were associated with CV events in the univariate analysis, only E/LA strain was still associated with adverse CV events after multivariate analysis. Furthermore, the addition of E/LA strain, but not LA strain, to a Cox model consisting of conventional clinical and echocardiographic parameters could cause an improvement in prediction of Scientific RepoRts | 6:17318 | DOI: 10.1038/srep17318 adverse CV events. Hence, the combination index, E/LA strain, might be more useful in predicting poor CV prognosis in AF patients than LA strain.
Several traditionally echocardiographic parameters including increased LA dimension, decreased LA appendage flow velocity, and proof of thrombi or spontaneous echo contrast were associated with an increased risk of stroke 3,4 . Recently, LA strain was reported to decrease proportionately with increasing CHADS 2 score and was an independent predictor of prior stroke 13,31 . Shih et al. found decreased LA strain and strain rate were independently associated with prior stroke, but E and E/E' were not 13 . The finding might indicate that stroke event was correlated better with LA-related parameters. However, this study was a cross-sectional study without long-term outcome data. In addition, it has also been proven that in patients with CHADS 2 scores ≦ 1, LA strain is an independent predictor of prior stroke, when adjusted for LA size, LVEF, and left ventricular mass 32 . In a recent observational study, LA strain provided  incremental value for embolism risk stratification over CHA 2 DS 2 -VASc score in patients with AF 30 . In the present study, we similarly found decreased LA strain was correlated with increased CHA 2 DS 2 -VASc score in the univariate analysis and further found decreased LA strain was significantly associated with subsequent stroke event even after adjustment for many important echocardiographic parameters. In addition, the addition of LA strain to a Cox model consisting of conventional clinical and echocardiographic parameters could cause an improvement in prediction of subsequent stroke event. Hence, impaired LA strain may be useful in predicting subsequent stroke event in AF patients. The underlying mechanism of the association between impaired LA strain and subsequent stroke event remains unclear. Previous studies have found there is an association among reduced LA strain, LA fibrosis, lower LA appendage flow velocity, and LA appendage thrombus 11,33 . We hypothesize that LA fibrosis diminishes LA compliance during the LA reservoir phase, which causes blood flow stasis in the left atrium. Therefore, reduced LA strain may contribute to an increased risk for subsequent stroke. However, E/LA strain was not associated with subsequent stroke event, which might also indicate that LA strain itself predicted future stroke event better than parameters of left ventricular filling pressure.

Conclusions
In patients with AF, E/LA strain and LA strain were useful in predicting adverse CV events and subsequent stroke event, respectively. Additionally, E/LA strain and LA strain could respectively provide incremental values for CV outcome and subsequent stroke outcome prediction over conventional clinical and echocardiographic parameters in AF.   Table 4. Incremental values of E/LA strain and LA strain in relation to CV events and subsequent stroke event. P value was based on the difference in − 2 log likelihood value compared with the previous model. CV: cardiovascular; LA: left atrial; LVEF: left ventricular ejection fraction.