Thromboembolism, notably stroke, is arguably the most feared complication in patients with nonvalvular atrial fibrillation. Huge efforts have been made in the past two decades to quantify and manage thromboembolic risk in such patients, resulting in substantial improvement in patient care and outcomes. However, more can be done to improve the stratification of thromboembolic risk in these patients. Of the several stroke risk schemes that have been described and compared, the CHADS ₂ score, evolved from the Stroke Prevention in Atrial Fibrillation investigators and the Atrial Fibrillation Investigators criteria, and promoted by the American Heart Association, American College of Cardiology, and European Society of Cardiology guidelines of 2006, has been the most widely accepted. CHADS ₂ is a method for predicting stroke risk in atrial fibrillation patients on the basis of clinical features: congestive heart failure, hypertension, age >75 years, diabetes, and prior stroke or transient ischemic attack. A comparison of 12 published stroke risk assessment schemes by the Stroke in Atrial Fibrillation Working Group found only a very modest predictive value for stroke with these schemes, with substantial and clinically relevant differences among the various schemes. More recently, a newer risk prediction scheme involving the addition of “clinically relevant non-major risk factors,” with the acronym the CHA ₂ DS ₂ -VASc, has been proposed and subsequently endorsed by the European Society of Cardiology in its recent new guidelines for the management of atrial fibrillation. This modification of the CHADS ₂ scoring scheme reflected an attempt to improve the discrimination of patients classified at “moderate risk” for stroke (particularly those with CHADS ₂ scores of 1), for whom opinion is most divided about the need for anticoagulation versus antiplatelet therapy. Despite attempts at improvement, the scoring systems that rely largely on clinical characteristics have low precision; the majority of patients recommended to receive anticoagulation therapy according to the predicted stroke risk will not develop stroke when treated with antiplatelet therapy alone, on the basis of the predictions from those same schemes. At the other end of the spectrum, there are some who are classified at low risk for stroke on the basis of current guidelines in whom atrial thrombi have been demonstrated.

In ≥75% of patients with atrial fibrillation who have stroke, left atrial (LA) thromboembolism is the culprit. Striking by its omission from all major prediction schemes, including the CHADS ₂ and CHA ₂ DS ₂ -VASc scores, therefore, is any parameter of LA size and function for predicting what after all are embolic events arising from that structure. The clinical characteristics associated with thromboembolic risk are mechanistically remote from the process of thrombus formation and thus are merely surrogates of local LA properties that predispose to thrombosis and embolization. It would appear logical, therefore, that incorporating descriptors of LA structure and function together with blood flow and rheologic characteristics, and possibly prothrombotic and anti-inflammatory factors, will improve predictive power.

The Stroke Prevention in Atrial Fibrillation investigators and the Atrial Fibrillation Investigators found moderate or severe left ventricular (LV) systolic dysfunction on echocardiography to be the only independent echocardiographic predictor of stroke in atrial fibrillation. Among the 12 most recognized risk stratification schemes in the literature for predicting stroke in patients with nonvalvular atrial fibrillation reported in 2008, LV systolic dysfunction was the only recognized independent echocardiographic risk factor. This situation still applies.

Although the CHADS ₂ score did not include LV systolic function as a predictive variable (the investigators did not have access to echocardiographic results), the 2006 guidelines of the American Heart Association, American College of Cardiology, and European Society of Cardiology nevertheless allowed LV ejection fraction ≤35% as a moderate risk factor for stroke. In the 2010 guidelines of the European Society of Cardiology, moderate or severe LV systolic dysfunction, arbitrarily defined as an ejection fraction ≤40%, was included as a surrogate for heart failure.

Many clinicians believe intuitively that LA size influences thromboembolic risk in atrial fibrillation, and many, despite the lack of definitive evidence in the literature, use this prejudice when deciding in borderline situations which patients should be offered anticoagulation therapy. So why is such evidence so glaringly lacking? Have we measured the wrong parameters, have we not measured the right parameters, or is there no influence of LA size and function on thromboembolic risk?

The initial attempts to assess the role of LA size in predicting thromboembolic risk in large studies of atrial fibrillation used the anteroposterior LA dimension measured by M-mode echocardiography. An early analysis by the Stroke Prevention in Atrial Fibrillation investigators of placebo-assigned patients found that LA diameter (and LV systolic dysfunction) was associated with the combined risk for ischemic stroke and systemic emboli. However, a subsequent analysis of a larger cohort of aspirin-treated patients from that study failed to confirm this association. A larger analysis of data from three major randomized atrial fibrillation trials (the Boston Area Anticoagulation Trial for Atrial Fibrillation, the Stroke Prevention in Atrial Fibrillation I study, and the Veterans Affairs Prevention in Atrial Fibrillation study) concluded that LA diameter by M-mode echocardiography did not predict stroke and at the same time confirmed the major influence of LV systolic dysfunction.

It is now recognized that there are substantial shortcomings in using the anteroposterior dimension to assess LA size, and biplane measurement of LA volume is the preferred method that is currently recommended in guidelines from the American Society of Echocardiography. The role of the LA size in thromboembolic risk prediction, although intuitively obvious, may not be as straightforward as initially envisaged. Clearly, LA enlargement is associated with many of the stroke risk factors in the CHADS ₂ scheme, such as hypertension, cardiac failure, and age (through associated comorbidities), and a larger atrium is expected to permit more blood flow stasis. However, in nonvalvular atrial fibrillation, the vast majority of the LA thrombi (91% in one study ) form in the appendage. This structure varies greatly in size and shape and may do so independently of the morphologic characteristics of the main atrial chamber. As such, mechanical and rheologic conditions favoring thrombus formation in the appendage may differ substantially from those found in the main atrial chamber. Indeed, a population-based study, albeit in patients in sinus rhythm, found that LA appendage flow velocities correlated poorly with global LA parameters. So, whether LA volume is an independent and clinically useful predictor of stroke risk in atrial fibrillation remains to be determined in large studies.

Despite the absence of LA parameters in contemporary stroke risk prediction schemes, the literature abundantly documents that characteristics of LA structure and function predict stroke risk in atrial fibrillation. Thus, the presence of spontaneous echo contrast in the LA and appendage, a marker of blood flow stasis best detected by transesophageal echocardiography, has long been recognized as a risk factor for thrombus formation and thromboembolic risk. Bridging the link between LV systolic dysfunction and stroke risk, and in keeping with the understanding that spontaneous echo contrast is an important final common pathway to thrombus formation, Rader et al. found in 524 patients with atrial fibrillation that LV ejection fraction <40%, age ≥75 years, and previous thromboembolism were independent predictors of LA spontaneous echo contrast detected by transesophageal echocardiography. In that study, LV ejection fraction <40% was the only multivariate predictor of the presence of LA thrombus. Linking clinical risk factors to stroke risk, the prevalence of LA spontaneous echo contrast and LA thrombus increased with increasing CHADS ₂ score. Wysokinski et al. similarly found that a number of risk factors in the CHADS ₂ score, atrial spontaneous echo contrast, and a few clinical descriptors of atrial fibrillation independently predicted the presence of LA appendage thrombus. A scheme incorporating these clinical and echocardiographic parameters predicted the presence of thrombus better than did the CHADS ₂ score. However, this proposal requires prospective validation.

Reduced blood flow velocity in the LA appendage has been found to be an important risk factor for thrombus formation. In the transesophageal echocardiographic substudy of the Stroke Prevention in Atrial Fibrillation III study of patients at moderate thromboembolic risk, LA thrombus, dense LA spontaneous echo contrast, low peak flow velocity in the LA appendage of ≤20 cm/sec, and complex aortic atheromatous plaque predicted an increase in the risk for thromboembolic events. In separate reports from this important substudy, age, LA and appendage flow dynamics, LA diameter measured by M-mode echocardiography, aortic atherosclerotic plaque, and plasma fibrinogen independently predicted the presence of dense spontaneous echo contrast, while LA appendage peak antegrade flow velocity <20 cm/sec and clinical risk factors independently predicted the presence of LA thrombus. Age, systolic blood pressure, sustained atrial fibrillation, ischemic heart disease, and LA area were independently associated with LA appendage peak antegrade flow velocity of <20 cm/sec. Of note, LA appendage cross-sectional area was not associated with stroke risk, suggesting that functional parameters might be more important than size in predicting thrombotic tendency, a finding supported by others.

This potentially important role of the LA appendage flow velocity in stroke prediction has been well corroborated. In a transesophageal echocardiographic examination of 500 patients with stroke in sinus rhythm or atrial fibrillation, the risk for LA appendage thrombus or spontaneous echo contrast, independently of rhythm, increased significantly at a LA appendage flow velocity of <55 cm/sec. When LA flow velocity was ≥55 cm/sec, there was only a minimal risk for thrombus and spontaneous echo contrast (negative predictive value, 99% and 100%, respectively). On multivariate analysis, LA appendage flow velocity was the strongest predictor of thrombus or spontaneous echo contrast.

In this issue of the Journal of the American Society of Echocardiography , Ayirala et al. provide further information relating clinical risk scoring, LA appendage function and LV systolic function to stroke risk in nonvalvular atrial fibrillation. In a consecutive series of 334 patients undergoing transesophageal echocardiography before cardioversion, the investigators detected thrombi in 15% of the patients. Higher CHADS ₂ score, larger LA volume index, and lower LV ejection fraction were significant predictors of LA appendage thrombus formation. LA appendage flow velocities were significantly lower in patients with LA appendage thrombus, but this association interestingly was lost in multivariate analysis, which is at odds with the results of other studies. Of note, no single echocardiographic parameter was able to distinguish between the presence or absence of atrial thrombus. The investigators make the point that LA appendage thrombus was not detected in patients with CHADS ₂ scores ≤1, LV ejection fractions >55%, and LA volume indexes <28 mL/m ² . However, interpretation of the data is difficult, because 95% of the patients were on some form of anticoagulation.

Unfortunately, LA appendage characteristics are best assessed by transesophageal echocardiography and less reliably by transthoracic imaging, making many of these parameters less accessible for screening purposes. Furthermore, changes in these parameters can occur over time and remain poorly understood, making revalidation of previous measurements necessary with repeat transesophageal intubation. In recent years, therefore, considerable attention has been given to finding alternative transthoracic echocardiographic measurements that may be clinically useful.

One such approach has been the use of two-dimensional speckle tracking of the LA myocardium to assess dynamic LA function. With this method, LA strain and strain rate may be measured in different segments of the atrium and at different phases of the cardiac cycle to represent the reservoir, conduit, and contractile components of LA function. This method of assessment has the major advantage over the use of Doppler tissue imaging by being angle independent, thus allowing more segments of the atrium to be analyzed. Methodology, feasibility, and the complex pathophysiologic correlations of LA strain and strain rate measurements have been reported in normal subjects and a number of pathologic conditions. Although initial studies are promising, feasibility for widespread clinical application remains to be demonstrated.

The study of Saha et al. reported in this issue of the Journal throws new light on this relatively new measure of LA function. The investigators used two-dimensional echocardiographic speckle tracking to measure global longitudinal strain during the reservoir phase as a potential predictor of stroke risk and cardiovascular outcomes in patients with nonvalvular atrial fibrillation. In this study of 36 patients with atrial fibrillation and 41 control subjects in sinus rhythm, global longitudinal LA strain was, as expected, lower in subjects with atrial fibrillation than in controls. In keeping with previous findings, indexed LA volume, LV ejection fraction, atrioventricular plane displacement, and pulmonary artery systolic pressure were related to global LA longitudinal strain by univariate analysis. On multivariate analysis, however, indexed LA volume and atrioventricular plane displacement were the only independent predictors of global LA longitudinal strain. On the basis of a CHADS2 score ≥2, LA longitudinal strain was related to the clinical characteristics reported to confer increased risk for stroke. Logistic regression analysis in the patients with atrial fibrillation found global LA strain to be the only echocardiographic variable associated with higher odds of having a CHADS ₂ score ≥2. Also, decreasing LA longitudinal strain was significantly associated with increasing CHADS ₂ stroke risk. The lack of a relationship between the indexed LA volume and the CHADS ₂ score was unexpected, but this could have been due to the relatively small numbers of patients studied. Indeed, there appeared to be a statistically insignificant trend for increasing indexed LA volume to be associated with increasing CHADS ₂ stroke risk.

The demonstration of a relationship between an echocardiographic parameter and the CHADS ₂ score is of interest but does not contribute practically to the clinical management of patients, because the CHADS ₂ score is clearly more convenient and less expensive to obtain than a relatively complex echocardiographic assessment. To be useful, an echocardiographic parameter needs to enhance stroke risk prediction beyond that already available from the commonly used prediction schemes.

The CHADS ₂ score has been previously shown to predict not just thromboembolism but also cardiovascular morbidity and mortality. In Saha et al. ’s study, incorporation of LA longitudinal strain and indexed LA volume to the CHADS ₂ score provided incremental value for the prediction of death and/or hospital admission over a median follow-up period of 394 days. Only 26 patients were studied in this way, and only 9 patients were hospitalized for unspecified cardiac events, with just 3 deaths. Thus, although the data are of interest, it would be premature to draw firm conclusions at this time about prognostic value and stroke risk prediction. Nevertheless, the results offer some promise for LA longitudinal strain to be one of the long-awaited missing links in stroke risk prediction using transthoracic echocardiography.

That there is an association between LA longitudinal strain and the CHADS ₂ score is, of course, not surprising, as LA reservoir function (which global longitudinal strain represents) has been reported to be related to age, LV hypertrophy and thus hypertension, and LV systolic function, factors found in the CHADS ₂ score. Regardless, this intriguing link between a global LA functional parameter and the clinical factors that predict stroke invites further investigation into possible mechanistic relationships.

The study of Shih et al. reported in this issue of the Journal examined such relationships by studying stroke history rather than clinical risk factors for stroke, yielding results that support the potential applicability of LA strain measurement for stroke risk prediction in atrial fibrillation. In 66 consecutive patients with permanent atrial fibrillation, LA peak longitudinal strain and peak strain rate in the reservoir phase and peak strain rate in the conduit phase were significantly lower in patients with previous stroke compared with those without previous stroke. On multivariate analysis, LA volume index, LV ejection fraction, and, notably, peak LA longitudinal strain and peak strain rate in the reservoir phase were independently associated with previous stroke.

One may ask why LA longitudinal strain should provide additional predictive information about stroke risk in atrial fibrillation beyond that available from LA volume and LV systolic function. At this juncture, answers can only be speculative. LA strain and strain rate relationships are complex, and our understanding of their implications is still in relative infancy. It has been known for many years that greater mitral annular motion from cardiac base to apex enhances reservoir function in healthy subjects and patients with dilated cardiomyopathy. In dog experiments, the early part of reservoir function is related to LA relaxation and the late part to the descent of the LV base toward the apex, representing systolic function. LA longitudinal strain in ventricular systole measured by two-dimensional echocardiographic speckle tracking is related to LA size and to LV systolic function. In the study of Wakami et al. , LA longitudinal strain during ventricular systole was significantly related to LV end-diastolic pressure and LV end-systolic volume index. This correlation between longitudinal strain and LV end-diastolic pressure was present in patients with and without impaired LV systolic function. Kurt et al. demonstrated that longitudinal LA strain measured during ventricular systole was lower in patients with diastolic heart failure.

Could the link between lower LA longitudinal strain and stroke risk simply reflect its association with larger LA volume? Geometrically, a larger LA volume should be accompanied by lower longitudinal strain for any change in volume during the reservoir phase. Clues can be gleaned from the reports published in this issue of the Journal . In the report of Saha et al. , indexed LA volume, together with atrioventricular plane displacement, was independently associated with global LA longitudinal strain, but global LA strain was the only echocardiographic variable associated with a CHADS ₂ score ≥2. Moreover, longitudinal strain exhibited a statistically significant gradient among three patient groups with increasing CHADS ₂ scores, while indexed LA volume did not. In the report of Shih et al. , LA longitudinal strain and indexed LA volume, together with other parameters, were independently associated with stroke. Thus, longitudinal strain appears to convey unique information not available from indexed LA volume alone.

LA volume is a static measurement, whereas LA longitudinal strain is a dynamic parameter that reflects the degree of change in LA length during filling. Perhaps more than the atrial size alone, this dynamic nature of the LA dimension may influence the stasis of blood lying alongside the LA endocardium and hence the tendency to atrial thrombosis.

Another mechanism by which LA longitudinal strain may carry unique information might be from its relation to myocardial fibrosis. A reduction in LA longitudinal strain has been associated with the presence of myocardial fibrosis, and in turn, LA fibrosis detected by late enhancement magnetic resonance imaging has been associated with stroke risk in patients with atrial fibrillation.

How strain and strain rate measurements in the atrial chamber relate to local behavior in the LA appendage remains unknown. Information on strain and strain rate in the LA appendage is scanty. Sevimli et al. studied this using transesophageal echocardiography and Doppler tissue imaging in patients in sinus rhythm, some of whom had native mitral stenosis, prosthetic mitral valves or heart failure. LA appendage strain and strain rate were significantly related to appendage emptying velocities, were lower in patients with spontaneous echo contrast, and were also lower in those with LA thrombus. In a transesophageal echocardiographic study of patients with stroke and atrial fibrillation by Ozer et al. , decreased LA appendage flow velocity, decreased LA wall velocity, increased LA appendage size, and less negative LA appendage tissue intensity predicted the presence of spontaneous echo contrast. However, LA appendage strain measurements offered no additional predictive benefit. Clearly, more work is required in this area.

Because screening with transesophageal echocardiography is impractical, a number of attempts have been made to characterize LA appendage structure and function by transthoracic evaluation. In a study by Moreira et al. , of 51 patients with acute ischemic neurologic conditions, all 7 patients with LA appendage thrombus or spontaneous echo contrast on transesophageal echocardiography had peak LA appendage emptying velocities on transthoracic echocardiography of <50 cm/sec. In the Comprehensive Left Atrial Appendage Optimization of Thrombus multicenter pilot trial of 118 patients with atrial fibrillation, apical LA appendage wall E velocity ≤0.7 cm/sec measured by transthoracic Doppler tissue imaging best indentified the patients with severe spontaneous echo contrast, sludge, or thrombus. An anterior LA appendage wall S velocity ≤5.2 cm/sec best identified the group of patients with sludge or thrombus. Both thrombi detected by transesophageal echocardiography were identified by second-harmonic transthoracic echocardiography and contrast. In another study, tissue Doppler velocity of the tip of the LA appendage measured from the transthoracic parasternal approach showed good correlation with LA emptying flow velocity measured by transesophageal echocardiography and was an independent predictor of LA appendage thrombus formation. More studies using echocardiographic equipment with contemporary technology are needed to establish widespread feasibility and determine if diagnostic sensitivities and specificities are superior and acceptable.

It is expected that the identification of characteristics at or close to the site of atrial thrombus formation in patients with nonvalvular atrial fibrillation will lead to greater accuracy in the prediction of thromboembolic risk and hence more precise management of such patients with anticoagulation therapy. We are all aware of many patients with long-standing atrial fibrillation who qualify for anticoagulation yet have never had stroke despite being on only aspirin. Indeed, the predictions from the various risk stratification schemes reflect this observation. The ability to reclassify patients from needing anticoagulation on the basis of current guidelines to not requiring it is highly desirable because of the not inconsiderable risk for bleeding associated with such treatment. This applies particularly to older patients and those with higher CHADS ₂ risk scores, who although at higher risk for stroke are at the same time at higher risk for bleeding on warfarin. The reverse also applies. Strokes uncommonly still occur in patients deemed on current criteria to be at low risk. Kleemann et al. described patients classified at relatively low thromboembolic risk on the basis of the CHADS ₂ score scheme who in fact were at higher than expected risk. In their patients with atrial fibrillation and low CHADS ₂ scores of 0 or 1, 3% had LA thrombus and 5% had dense spontaneous echo contrast. In these patients, independent predictors of the presence of dense spontaneous echo contrast were LV ejection fraction ≤40% and LA dimension ≥50 mm. The 3% prevalence of atrial thrombus is identical to that found by Rader et al. in their cohort of patients with atrial fibrillation and CHADS ₂ scores of 0. Finally, the identification of patients in atrial fibrillation and on anticoagulation who remain at risk for stroke will also be valuable so that further intervention may be considered.

Technology in echocardiography has developed much since the era of the large trials of warfarin in atrial fibrillation of the late 1980s and early 1990s. Aided by better equipment and armed with new clues regarding prospective atrial and ventricular predictors, including those in the current issue of the Journal , the time is ripe for a renewed, collective effort to identify for clinical application the mechanistic as opposed to the clinical promoters of atrial thrombus formation and embolization in atrial fibrillation.

The task ahead will not be easy. On practical grounds, the widespread adoption of such an approach will need the use of transthoracic echocardiographic measurements with proven feasibility in the majority of patients and with acceptable reproducibility. Clinically useful information with sufficiently robust positive and negative predictive values will need to come from large population studies of patients with atrial fibrillation who are not anticoagulated, or from studies of large cohorts of patients in major atrial fibrillation trials not on anticoagulation therapy. The increasing adoption of anticoagulation guidelines will necessarily affect the nature of such populations and introduce a selection bias. The advent of a new era of more convenient and safer nonwarfarin anticoagulants such as dabigatran and apixaban, with more such agents to follow, is expected to lead to a greater uptake of anticoagulation in atrial fibrillation population, particularly in the group at moderate risk for stroke, which will change the landscape in which such studies are performed. As a consequence, the determination of the “natural history” of thromboembolic risk in patients who are not on anticoagulation will become even more difficult than it is now.

In the realm of thromboembolic risk prediction in atrial fibrillation, echocardiographic research arguably has been lagging behind, with the result that, contrary to expectation, no echocardiographic parameter other than LV systolic dysfunction is included in contemporary stroke risk prediction. The time to catch up has arrived, and in timely fashion, the three reports of Ayirala et al. , Saha et al. , and Shih et al. in this issue of the Journal offer encouragement to press ahead.