Author + information
- Received October 12, 2017
- Revision received December 12, 2017
- Accepted December 28, 2017
- Published online February 14, 2018.
- Flemming J. Olsen, MDa,b,∗ (, )
- Rasmus Møgelvang, MD, PhDa,b,
- Gorm B. Jensen, MD, DMSca,
- Jan S. Jensen, MD, PhD, DMSca,b,c and
- Tor Biering-Sørensen, MD, PhDa,b,d
- aCopenhagen City Heart Study, Copenhagen, Denmark
- bDepartment of Cardiology, Herlev & Gentofte Hospital, University of Copenhagen, Copenhagen, Denmark
- cInstitute of Clinical Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark
- dDepartment of Medicine, Cardiovascular Medicine Division, Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts
- ↵∗Address for correspondence:
Dr. Flemming J. Olsen, Department of Cardiology, Herlev & Gentofte Hospital, University of Copenhagen, Niels Andersens Vej 65, 2900 Hellerup, Denmark.
Objectives This study sought to investigate whether left atrial (LA) functional measures predict atrial fibrillation (AF) in the general population.
Background Increasing evidence suggests LA functional measures are predictors of AF in several patient groups.
Methods In a community-based cohort study, approximately 2,000 individuals underwent a transthoracic echocardiogram. Conventional echocardiographic measures and extended LA measures, including the minimal and maximal LA volumes (LAVmin and LAVmax, respectively) and left atrial emptying fraction (LAEF), were performed. The endpoint was incident AF, and participants with known AF were excluded, which left 1,951 for inclusion.
Results Over 11.0 years of follow-up, 184 (9.4%) developed AF. Those who developed AF had significantly larger LA volumes and lower LAEF than participants free of AF. These LA measures were univariable predictors of AF (LAVmax hazard ratio [HR]: 1.10 [95% confidence interval (CI): 1.08 to 1.12] per 1-ml increase, p < 0.001; LAVmin HR: 1.14 [95% CI: 1.12 to 1.16] per 1-ml increase, p < 0.001; LAEF HR: 1.03 [95% CI: 1.02 to 1.04] per percent decrease, p < 0.001). All LA measures remained predictors independent of clinical risk scores, with LAVmin providing the highest C-statistics when added to these risk scores (C-statistic for CHADS2 0.728 vs. CHADS2 + LAVmin 0.778; C-statistic for CHARGE-AF 0.815 vs. CHARGE-AF + LAVmin 0.830). However, hypertension modified the relationship between the measures of LA function (both LAVmin and LAEF) and risk of AF (p for interaction < 0.001), which was not the case for LAVmax (p = 0.22). The measures of LA function mainly provided prognostic information regarding risk of AF in participants without hypertension. Even when we restricted our analysis to individuals without hypertension and nondilated LA (LAVmax<34 ml/m2), the LAVmin and LAEF remained significantly independent predictors of AF after multivariable adjustments (LAVmin HR: 1.12 [95% CI: 1.01 to 1.24], p = 0.028, and LAEF HR: 1.03 [95% CI: 1.00 to 1.06], p = 0.021, respectively).
Conclusions LA functional measures predict AF in the general population and provide prognostic information incremental to clinical risk scores. In individuals without hypertension and nondilated LA, these measures indicate an increased risk of AF.
Atrial fibrillation (AF) is the most common cardiac arrhythmia, afflicting approximately 3% of the general population, and the prevalence of this arrhythmia is expected to rise steeply (1). AF significantly increases the risk of stroke and heart failure, and development of AF is associated with poor outcome in several patient populations (2,3). However, the arrhythmia is manageable, and a significant risk reduction in stroke is possible with antiplatelet and anticoagulant drugs according to established guidelines (1). Thus, early detection of AF is important for timely stroke prevention and overall improvement in prognosis.
Several echocardiographic measures have been proposed as predictors of AF, which could help identify patients at high risk of AF. Recently, studies from high-risk patients, that is, patients with cryptogenic stroke, have unveiled the potential of functional left atrial (LA) measures such as the minimal LA volume (LAVmin) and the LA emptying fraction (LAEF) for the prediction of AF (4). A few imaging studies on community-based cohorts have also revealed these measures to be reliable predictors of incident AF (5,6). Thus, there could be prognostic potential to be gained from other measures than the maximal LA volume (LAVmax), which is the only LA measure applied clinically (7). However, the long-term prognostic potential of extensive LA volume measurements by echocardiography for prediction of AF in a large-scale general population is unknown. Additionally, whether LA measures have value beyond established clinical risk scores is unknown. Finally, no community-based cohort studies have investigated the influence of hypertension on the predictive ability of LA measures. This is relevant because hypertension is a well-established risk factor for AF, and studies on hypertensive individuals have found that both structural and functional measures of the LA are largely influenced by the presence of this state (8). Thus, the same predictors of AF might not apply to these individuals, and this is important given the high prevalence of hypertension in the general population.
Therefore, we sought to determine whether LAVmin and LAEF are superior predictors of AF compared with LAVmax and whether these measures add prognostic value beyond clinical risk score models. Second, we wanted to assess whether these measures could predict incident AF in participants with structurally normal LA (LAVmax<34 ml/m2). Finally, we wanted to investigate whether hypertension modifies the relationship between LA measures and risk of AF.
The present study examined participants who were enrolled as part of a large-scale community-based cohort study, the 4th Copenhagen City Heart Study. Information on the overall Copenhagen City Heart Study has been described previously (9,10), as has the echocardiographic study subcohort (11). Individuals from this cohort who were free of AF and who had a thorough transthoracic echocardiogram performed in 2002 to 2003 (as part of the fourth round of examination in the overall Copenhagen City Heart Study) were included in this echocardiographic substudy (n = 1,951).
The study complied with the Second Helsinki Declaration (and recent amendments) and was approved by a regional ethics committee (identification number: H-KF 01-144/01 31104) and the Danish Data Protection Agency (journal number 2007-58-0015), and informed consent was obtained from all participants.
Information on the echocardiographic examinations and measurements is described in detail in the Online Appendix. Of note, LA volumes were measured by the biplane area-length method at end systole (LAVmax) and end diastole (LAVmin). LAEF was calculated as the fractional change between these 2 measures (LAVmax − LAVmin) / LAVmax.
Participants were included from the fourth round of examination in 2002 to 2003 and followed up until April 2013, the time of AF event, or the time of death. The outcome was new-onset AF, and endpoints were obtained by International Classification of Disease (ICD-10) codes from the Danish Board of Health’s National Patient Registry with a follow-up rate of 100%. The ICD-10 code was I48.9 and included all types of AF (paroxysmal, persistent, and permanent AF and atrial flutter).
STATA data analysis and statistical software (Stata/SE version 12.0, StataCorp, College Station, Texas) was used for statistical calculations. Continuous variables exhibiting gaussian distribution were compared between the groups by Student t test and expressed as mean ± SD. Those not showing gaussian distribution were compared by Wilcoxon rank sum test and expressed as median (interquartile range). The chi-square test was applied for binary and categorical variables and expressed as percentages. A p value ≤0.05 in 2-tailed tests was considered statistically significant.
We performed reproducibility analysis by examining intraobserver and interobserver variability for LAVmax, LAVmin, and LAEF in 19 randomly selected participants from this cohort. We calculated coefficients of variation (CVs) and bias coefficients (mean difference ± SD) and visually displayed the variability with Bland-Altman plots (Online Figures 1 to 6).
We tested for collinearity in our multivariable model 3 (including confounders identified within the cohort) by calculating the variance inflation factor (VIF) with a threshold VIF of 5.
Univariable Cox regression was performed to correlate clinical, biochemical, and echocardiographic findings to the endpoint. Univariable predictors were incorporated into multivariable Cox regression models to adjust for potential confounders and for calculation of adjusted hazard ratios (HRs). We also performed multivariable Cox regression with the inclusion of the CHADS2 score (a score that assigns 1 point for presence of congestive heart failure, hypertension, age 75 years or older, and diabetes mellitus and 2 points for a history of stroke or transient ischemic attack) (12) and the CHARGE-AF (Cohorts for Heart and Aging Research in Genomic Epidemiology–Atrial Fibrillation) score (13) (see the Online Appendix for the CHARGE-AF calculation).
We also investigated other measures of LA function (mitral A-wave and global a′) to examine whether these parameters were independent predictors of the clinical risk score models. Subdistributional HRs were also calculated by competing risk Cox proportional hazard regression to test whether handling all-cause mortality as a competing risk influenced the relationship between LA measures and AF outcome. We furthermore performed interaction analysis to test whether hypertension modified the relationship between LA measures and outcome.
Harrell C-statistics were calculated from univariable Cox regression for all LA measures to compare predictive capability, as well as from multivariable Cox regression models for clinical predictors and risk models of AF, and for LA measures added to these predictors and models.
Rate of events based on LA measures are depicted in Figures 1, 2⇓⇓, and 3 as incidence rates, defined as events per person-time at risk of AF. The same is done in Online Figures 7, 8, and 9, but with the population stratified as hypertensive versus normotensive. These incidence rate curves were constructed by the use of Poisson models.
Of the 1,951 participants included in this study, 184 (9.4%) developed the outcome of new-onset AF. Follow-up was 100% during a median follow-up period of 11.0 years (interquartile range: 9.8 to 11.2 years).
Baseline characteristics for the entire population and baseline characteristics based on outcome of AF are portrayed in Table 1. The majority (57%) of participants in this study were women, and the mean age was 59 years. A substantial proportion (44%) had hypertension, and the mean left ventricular (LV) ejection fraction was 60%.
Participants who developed AF had larger LA dimension, larger LA volumes, and lower LAEF than those who remained free of AF (LA dimension: 3.7 vs. 3.4 cm; LAVmin: 13 vs. 9 ml; LAVmax: 23 vs. 19 ml; LAEF: 45 vs. 51%; p < 0.001 for all).
Reproducibility and collinearity
In intraobserver analysis, we found no difference between LAVmax and LAVmin (CV = 10.7% for both). LAEF had the lowest intraobserver variability (CV = 9.0%). In interobserver analysis, we found that LAVmax had a lower variability than LAVmin (CV = 24.6% vs. 29.3%, respectively), with LAEF still showing the lowest variability (CV = 19.9%).
The bias coefficients were as follows: LAVmax intraobserver: −1.85 ± 4.01 ml/m2; LAVmax interobserver: 1.41 ± 8.79 ml/m2; LAVmin intraobserver: 0.26 ± 1.89 ml/m2; LAVmin interobserver: 0.76 ± 5.12 ml/m2; LAEF intraobserver: −3.25 ± 4.75%; LAEF interobserver: 0.25 ± 10.16% (Online Figures 1 to 6). No variables in multivariable model 3 (Table 2) exhibited a VIF > 5.
LA measures in relation to clinical risk models
Both the CHADS2 and CHARGE-AF risk scores were strong predictors of AF in the population (HR: 2.22 [95% confidence interval (CI): 1.95 to 2.51], p < 0.001 and 2.32 [95% CI: 2.05 to 2.63], p < 0.001, respectively). In the overall population, all LA measures were independent predictors of AF after adjustment for both risk models (Table 2). Both LAVmin and LAVmax increased the Harrell C-statistics when added to both models, whereas LAEF only increased the Harrell C-statistic significantly when added to the CHADS2 score. The results did not differ in competing risk analysis (Table 3). In all models, the LAVmin provided the highest Harrell C-statistics (Table 2).
Predictive value of LA measures in other models
Several clinical, paraclinical, and echocardiographic parameters were found to be significant predictors of AF in univariable Cox regression (Table 2, model 3). Multivariable Cox regression encompassing these clinical, paraclinical, and echocardiographic variables showed that for the LA variables, only the LA volumes remained significant predictors of outcome (LAVmin HR: 1.04 [95% CI: 1.00 to 1.08], p = 0.033; LAVmax HR: 1.03 [95% CI: 1.00 to 1.06], p = 0.027) (Table 2). However, competing risk analysis revealed that subdistributional HRs for LAVmax differed from LAVmax in multivariable Cox regression, and LAVmax just failed to attain a statistically significant predictive effect on AF when correcting for all-cause mortality as a competing risk (subdistributional HR: 1.03 [95% CI: 1.00 to 1.06]; p = 0.06). The same result was seen for the LAVmin; however, in more advanced models that accounted for hypertension, which modified the relationship between LAVmin and AF, LAVmin remained an independent predictor of AF. Furthermore, LAEF became a significant predictor of AF in competing risk analysis in this advanced model (modified model 3, Table 3).
Harrell C-statistics showed that LAVmin provided the highest predictive value (Table 2) compared with the other LA measures. No LA measure provided a significant increment in C-statistics to that of model 3, which yielded a very high C-statistic of 0.859 (Table 2).
Hypertension and LA measures
Interaction analysis revealed that hypertension modified the relationship between both LAVmin and LAEF and the endpoint of AF (p < 0.001), which was not the case for LAVmax (p = 0.22). The same interaction was found in the competing risk analysis.
When the entire population was stratified according to hypertension, it became evident that none of the LA measures were significant predictors among hypertensive individuals after multivariable adjustment for confounders identified within this cohort (Table 4). However, both LAVmax and LAVmin were independent predictors after adjustment for both the CHADS2 and CHARGE-AF risk scores in the subgroup of hypertensive individuals (Table 4). They also significantly increased the Harrell C-statistics in both of these models. In the subgroup of participants with hypertension and nondilated LA (defined as LAVmax < 34 ml/m2), LAVmax was the only LA measure that proved to be a predictor when adjusted for the CHADS2 score (HR: 1.06 [95% CI: 1.03 to 1.09]; p < 0.001) and CHARGE-AF score (HR: 1.04 [95% CI: 1.00 to 1.07]; p = 0.027), but it significantly increased the Harrell C-statistic only when added to the CHADS2 score (C-statistic 0.641 vs. 0.688 for CHADS2 and CHADS2 + LAVmax, respectively), not the CHARGE-AF score.
For normotensive individuals, all LA measures were independent predictors when adjusted for the CHADS2 and CHARGE-AF risk scores, and all increased the Harrell C-statistics compared with the CHADS2 scores, but none increased the C-statistics compared with the CHARGE-AF score. When we restricted our analysis to normotensive individuals and nondilated LAs, all LA measures remained independent predictors of AF after adjustment for the clinical risk scores; however, only LAVmin and LAEF significantly increased the Harrell C-statistics (0.606 for CHADS2; 0.711 for CHADS2 + LAVmin; and 0.681 for CHADS2 + LAEF). In our multivariable model that adjusted for confounders identified within this cohort, both LAVmin and LAEF remained independent predictors of AF (HR: 1.13 [95% CI: 1.03 to 1.25], p = 0.010 and HR: 1.03 [95% CI: 1.00 to 1.05], p = 0.036, respectively), whereas LAVmax did not (Table 4). Even when we restricted our analysis to normotensive individuals without evident enlargement of the LA, these 2 measures remained independent predictors of incident AF (LAVmin HR: 1.12 [95% CI: 1.01 to 1.24], p = 0.028 and LAEF HR: 1.03 [95% CI: 1.00 to 1.06], p = 0.021). No other echocardiographic measures remained independent predictors in these models.
Incidence rate curves (Figures 1 to 3) showed that with increasing LA volumes (both LAVmin and LAVmax), there was a linearly increased risk of AF. The inverse pattern was seen for LAEF; with decreasing percentile of LAEF, there was a linearly increased risk of AF. When portraying participants as hypertensive versus normotensive (Online Figures 7 to 9), it became apparent that normotensive individuals had much steeper incidence rate curves and higher incidence risk per increase in milliliters and decrease in percentages for volumes and LAEF, respectively.
When we only looked at hypertensive individuals, we found that those who developed AF had a higher left ventricular mass index (LVMI), and more of them had LV hypertrophy compared with hypertensive individuals who did not develop AF (LVMI 105 vs. 92 g/m2, p < 0.001 and LV hypertrophy 42 vs. 28%, p = 0.007).
Predictive value of other atrial measures
Univariable Cox regression revealed that the mitral A-wave was a predictor of AF in the overall population; however, it did not remain an independent predictor in the overall population after adjustment for the CHADS2 and CHARGE-AF risk scores in multivariable Cox regression, nor when participants were stratified according to hypertensive status. The global a′ was not a univariable predictor of AF in the overall population nor in normotensive individuals, but in hypertensive individuals, it was an independent predictor of AF after adjustment for both CHADS2 and CHARGE-AF scores (HR: 1.25 [95% CI: 1.13 to 1.39], p < 0.001 and HR: 1.28 [95% CI: 1.16 to 1.42], p < 0.001 respectively), and it significantly increased the Harrell C-statistic in both models (0.655 for CHADS2 vs. 0.702 for CHADS2 + global a′ and 0.727 for CHARGE-AF vs. 0.757 for CHARGE-AF + global a′, p < 0.05 for both).
In this large-scale cohort study of prospectively enrolled participants from the general population, we made several important findings: 1) we found all LA measures to be significant independent predictors of incident AF in the whole population and to add valuable prognostic information to clinical risk models (both the CHADS2 and CHARGE-AF), but in more advanced models, only LAVmin remained an independent predictor of AF; 2) in individuals without hypertension, only LA functional measures were independent predictors in all regression models; 3) in this subgroup, LA functional measures can even predict AF in individuals with structurally normal LA (defined as LAVmax<34 ml/m2); and 4) in those with hypertension, both the LAVmin and LAVmax were independent predictors of AF, however, only LAVmax indicated an increased risk of AF in those with a nondilated LA. To the best of our knowledge, this is the largest echocardiographic study to examine the prognostic ability of LA measures for the prediction of AF in the general population.
LAVmin vs. LAVmax
Only a few community-based cohort studies have evaluated the prognostic potential of LA measures to predict incident AF, and none have had as long a follow-up as in this study. A decade ago, Tsang et al. (14) established that LAVmax was superior to the LA diameter for prediction of incident AF in an elderly community, and the most recent expert consensus statement has recognized the LA diameter as an outdated measure (7). More recent results from Fatema et al. (5) examined the potential of the LAVmin and found that this measure was actually a superior prognostic marker compared with LAVmax but was subject to greater interobserver and intraobserver variability than the LAVmax. Because the LAVmin was only marginally better than the LAVmax, the higher reproducibility of the LAVmax favored this measure clinically. However, that study was smaller than our population (n = 574), with fewer events (n = 30, or 5%), was restricted to elderly citizens (age > 65 years), and had a shorter follow-up than our study. Our findings suggest similar intraobserver variability between the LAVmax and LAVmin, but a slightly higher interobserver variability for LAVmin than LAVmax. Interestingly, we found LAEF to be the most reproducible parameter. The fact that LAVmax was not an independent predictor of AF in subdistributional Cox regression in our study might favor the application of LAVmin in the clinic. This measure is only marginally time-consuming to perform, and given that LA planimetry is already a standardized method, the introduction of a new LA volume is realistic compared with other rather complicated measures of LA function (e.g., measures of electromechanical dyssynchrony by tissue Doppler imaging or LA speckle tracking ). The only other large-scale community-based cohort study to examine LA measures is MESA (Multi-Ethnic Study of Atherosclerosis) (6); however, this was by cardiac magnetic resonance imaging. Among 6,000 individuals, 517 were included in a substudy, in which both LAVmax and LAEF were found to be independent prognostic markers of AF. Unfortunately, the authors did not report the prognostic ability of LAVmin.
However, imaging studies in patients with myocardial infarction (16), aortic valve stenosis (17), and cryptogenic stroke (4) have made similar findings to ours, and increasing evidence suggests that a large proportion of patients who develop AF do not have an enlarged LA by the LAVmax (18). Findings from the ENGAGE-AF TIMI 48 (Effective Anticoagulation With Factor Xa Next Generation in Atrial Fibrillation – Thrombolysis in Myocardial Infarction 48) trial (16) actually suggest that more than one-third of AF patients do not have enlarged LA, and for paroxysmal AF, this proportion is even higher (52%). In addition, this study showed that more patients had impaired functional measures, which suggests that LA functional measures are affected before LA enlargement by the LAVmax becomes evident. An invasive hemodynamic study by Appleton et al. (19) showed that the LAVmin is more closely related to all invasive measures of LV filling pressure than LAVmax, and this might contribute to remodeling of the LAVmin before the LAVmax, thereby explaining why the LAVmin is a superior predictor of AF compared with LAVmax.
Hypertension and LA measures
Hypertension modified the relationship between LA functional measures and AF outcome, leaving these measures as nonsignificant predictors of AF in participants with hypertension. However, it is noteworthy that the LAVmax was also not a predictor after multivariable adjustments and after competing risk analysis. These findings might reflect that hypertension itself causes LA remodeling because of the increased afterload conditions (20), which makes it difficult to recognize the minor LA structural changes that are induced by underlying paroxysmal AF. This notion is supported by the fact that the hypertensive participants who experienced the outcome of AF had more pronounced LV remodeling than those who did not develop AF (LVMI: 105 vs. 92 g/m2, p < 0.001; LV hypertrophy: 42 vs. 28%, p = 0.007). This, however, begs the question as to which echocardiographic predictors should then be used in hypertensive patients for prediction of AF. One option that must be considered is that the LAEF encompasses both active and passive properties of the LA (21), and because about one-half of patients with hypertension have impaired relaxation of the LV (22), impairment in passive filling of the LV might not become apparent by the LAEF. This is because an impaired passive filling of the LV would be met by a compensatory increase in active filling, and this would go unnoticed in our LA functional measures. Analyses of other measures of LA function in our study, however, suggest that decreased atrial contraction by the global a′ could be a valuable parameter with respect to predicting AF in hypertensive individuals. This might reflect that the atrial reserve becomes exhausted and that patients are progressing in diastolic dysfunction beyond impaired relaxation. By extension, isolated LAEF might not have value in these patients, but passive and active LAEF could be of potential value. Furthermore, several recent studies have examined the prognostic value of LA speckle tracking and the application of LA functional measures derived from this modality. This technique might hold value pending results of further studies (23).
Because we obtained endpoints from registries using ICD-10 codes, we cannot state how rigorously the participants were monitored for AF. Although diagnostic misclassifications can occur in registries, we have previously found a high sensitivity and specificity for the AF diagnosis through validation of random subjects (24).
We only measured the LAVmin and LAVmax in our study, not the LA volume at the p-wave. LA volume at the p-wave would have allowed us to measure both passive and active LAEF, which could potentially have provided additional prognostic information and insight as to what type of diastolic dysfunction accompanies an increased risk of AF.
The Cox regression model is biased by overfitting because of the small number of AF events, particularly in our subgroup analyses, which warrants caution for the interpretation. However, this approach can be acceptable when the focus is on controlling for confounders rather than building prediction models (25).
Functional measures of the LA, the LAVmin and LAEF, are independent predictors of incident AF in the general population, which was not the case for the LAVmax. This particularly applies to individuals without hypertension, and even in people without an enlarged LA, these measures indicate an increased risk of AF, which goes undetected by the LAVmax.
COMPETENCY IN MEDICAL KNOWLEDGE: Left atrial remodeling poses an increased risk of atrial fibrillation, and early recognition of ongoing remodeling might help to select patients in need of rhythm monitoring. Extended left atrial measures such as the minimal left atrial volume and left atrium emptying fraction could facilitate this process before remodeling is recognized by the currently applied maximal left atrial volume.
TRANSLATIONAL OUTLOOK: Prospective trials in individuals with continuous rhythm monitoring are needed to determine how extended left atrial volumes and more advanced measures of left atrial function, that is, speckle tracking–derived measures, are affected in the remodeling process and how this relates to an increased risk of atrial fibrillation, as well as whether these more advanced echocardiographic measures are needed to risk stratify patients.
The Copenhagen City Heart Study was financially supported by the Danish Heart Foundation, and the echocardiographic substudy of the 4th round of examination was supported by the Lundbeck Foundation. Dr. Olsen was funded by grants from the Herlev & Gentofte Hospital’s Research Council and the P. Carl Petersen Foundation. All authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- atrial fibrillation
- confidence interval
- coefficient of variation
- hazard ratio
- International Classification of Diseases, 10th Revision
- left atrium
- left atrium emptying fraction
- maximal left atrial volume
- minimal left atrial volume
- left ventricle
- left ventricular mass index
- variance inflation factor
- Received October 12, 2017.
- Revision received December 12, 2017.
- Accepted December 28, 2017.
- 2018 American College of Cardiology Foundation
- Ruwald A.-C.H.,
- Bloch Thomsen P.E.,
- Gang U.,
- Jørgensen R.M.,
- Huikuri H.V.,
- Jons C.
- Anter E.,
- Jessup M.,
- Callans D.J.
- Biering-Sørensen T.,
- Christensen L.M.,
- Krieger D.W.,
- et al.
- Habibi M.,
- Samiei S.,
- Ambale Venkatesh B.,
- et al.
- Xu T.-Y.,
- Sun J.P.,
- Lee A.P.-W.,
- et al.
- Appleyard M.,
- Hansen A.T.,
- Schnohr P.,
- Jensen G.,
- Nyboe J.
- Christophersen I.E.,
- Yin X.,
- Larson M.G.,
- et al.
- Tsang T.S.M.,
- Abhayaratna W.P.,
- Barnes M.E.,
- et al.
- Sarvari S.I.,
- Haugaa K.H.,
- Stokke T.M.,
- et al.
- Gupta D.K.,
- Shah A.M.,
- Giugliano R.P.,
- et al.,
- for the Effective Anticoagulation With Factor Xa Next Generation in AF–Thrombolysis in Myocardial Infarction 48 (ENGAGE AF-TIMI 48) Echocardiographic Study Investigators
- Pagola J.,
- González-Alujas T.,
- Flores A.,
- et al.
- Appleton C.P.,
- Galloway J.M.,
- Gonzalez M.S.,
- Gaballa M.,
- Basnight M.A.
- Hoit B.D.
- Schillaci G.,
- Pasqualini L.,
- Verdecchia P.,
- et al.
- Olsen F.J.,
- Bertelsen L.,
- de Knegt M.C.,
- et al.
- Mukamal K.J.,
- Tolstrup J.S.,
- Friberg J.,
- Jensen G.,
- Grønbaek M.