Atrial fibrillation echo analysis. CI indicates confidence interval; LV, left ventricle.
Atrial Fibrillation Investigators: Atrial Fibrillation, Aspirin, Anticoagulation Study; European Atrial Fibrillation Study; Stroke Prevention in Atrial Fibrillation Study; Boston Area Anticoagulation Trial for Atrial Fibrillation Study; Canadian Atrial Fibrillation Study; Veterans Affairs Prevention in Atrial Fibrillation Study. Echocardiographic Predictors of Stroke in Patients With Atrial FibrillationA Prospective Study of 1066 Patients From 3 Clinical Trials. Arch Intern Med. 1998;158(12):1316-1320. doi:10.1001/archinte.158.12.1316
Clinical features that consistently predict ischemic stroke in patients with nonvalvular atrial fibrillation have been identified, while echocardiographic risk factors are less well defined.
To determine whether the results of transthoracic echocardiography add independent information to the clinical risk factors for stroke in patients with atrial fibrillation.
Transthoracic echocardiographic findings and clinical features from 1066 patients with atrial fibrillation assigned to placebo or control in 3 randomized trials (Boston Area Anticoagulation Trial for Atrial Fibrillation, Stroke Prevention in Atrial Fibrillation I Study, and Veterans Affairs Prevention in Atrial Fibrillation Study) were correlated with subsequent ischemic stroke by multivariate analysis.
The mean±SD age of patients was 67±10 years, 78% were men, 55% had a history of hypertension, 19% had a history of diabetes, 7% had a previous transient ischemic attack or stroke, and 27% had a history of heart failure. During a mean follow-up of 1.6 years, 78 ischemic strokes occurred (annual rate, 4.7%). Moderate to severe left ventricular systolic dysfunction shown via 2-dimensional echocardiography was a strong independent predictor of stroke (relative risk, 2.5; P<.001) in the 1010 patients in whom echocardiographic values for left ventricular function were available. Left atrial diameter by M-mode echocardiography did not predict stroke (relative risk, 1.02/mm; P=.10). Of 163 patients categorized as low risk based on clinical features (annual stroke rate, 0.8%; 95% confidence interval, 0.2%-3.0%), 10 had moderate to severe left ventricular dysfunction shown via 2-dimensional echocardiography and a 9.3% per year risk of stroke (95% confidence interval, 1.3%-66%). Conversely, 728 of the 847 patients at high risk for stroke based on clinical criteria had normal or mildly abnormal left ventricular function; their stroke rate was 4.4% (95% confidence interval, 3.4%-5.8%).
Left ventricular systolic dysfunction shown via 2-dimensional transthoracic echocardiography independently predicts risk of stroke in patients with atrial fibrillation. Echocardiography may prove most useful in a small group of patients who have a low risk of stroke according to clinical factors.
WE PREVIOUSLY reported clinical predictors of ischemic stroke in patients with nonvalvular atrial fibrillation (AF) based on the pooled results of 5 randomized clinical trials.1 These predictors were increasing age, previous stroke or transient ischemic attack (TIA), history of hypertension, or diabetes. Three of these trials2- 4 collected transthoracic echocardiographic data at study entry. In this analysis, we pooled individual patient data from these 3 trials to identify the echocardiographic predictors of stroke and to assess whether echocardiography enhanced risk stratification compared with using clinical predictors alone.
Echocardiographic data from individual patients obtained at entry into 3 clinical trials were assembled in a common database, as previously reported.1 Patients were included if they were assigned to placebo or control in the Stroke Prevention in Atrial Fibrillation (SPAF) I study,2 the Veterans Affairs Stroke Prevention in Nonrheumatic Atrial Fibrillation (SPINAF)3 study, or the Boston Area Anticoagulation Trial for Atrial Fibrillation (BAATAF).4 In the BAATAF study, 46% of patients in the control group took various doses of aspirin.4 The primary outcome for this analysis was ischemic stroke, defined as the abrupt onset of a focal neurologic deficit that lasted more than 24 hours. Hemorrhages were excluded by neuroimaging in most cases. No attempt was made to distinguish cardioembolic stroke from noncardioembolic stroke.
For determination of left atrial diameter, all studies used either the 2-dimensional (2-D) parasternal long axis or short axis view to guide acquisition of the M-mode left atrial tracing. The BAATAF study measured the inside edges rather than the leading- to inside-edge method used by the other 2 studies. To reconcile the 2 methods, a random set of 19 echocardiograms from the BAATAF study were independently interpreted using the leading- to inside-edge method by 1 echocardiographer from each study. For each individual echocardiographer the reinterpreted values were statistically significantly higher than the original BAATAF values (for each, P=.004, P=.004, and P<.001, respectively). The mean absolute difference was 5.5 mm. Unless otherwise indicated, all analyses of left atrial diameter in this study included an adjustment for this systematic difference. The BAATAF results were also systematically different from the other 2 studies for the variables intraventricular septal thickness, left ventricular posterior wall thickness, left ventricular end-diastolic volume, and left ventricular mass. It was assumed that this was also due to a difference in measurement technique, and analyses of these variables were thus adjusted for study as well.
Analysis of M-mode echocardiograms was otherwise based on standard criteria.5 Left ventricular muscle mass was calculated as follows6:
where IVS indicates interventricular septal thickness at end diastole (centimeters); LVPW, left ventricular posterior wall thickness at end diastole (centimeters); and LVEDD, left ventricular end-diastolic dimension (centimeters). Left ventricular systolic function was categorized qualitatively by 2-D echocardiography in all 3 studies, when studies were technically adequate. The SPAF study assessed "overall" and "focal" left ventricular wall motion as normal, mildly abnormal, and moderately to severely abnormal. The SPINAF study assessed "global" wall motion as normal, mildly abnormal, moderately abnormal, or severely abnormal, and "regional" wall motion, in 3 anatomical regions, as normal, hypokinetic, akinetic, or dyskinetic. The BAATAF study assessed 6 separate regions of the left ventricle as normal, hypokinetic, akinetic, or dyskinetic. For the current analysis left ventricular systolic function was defined as mildly abnormal as follows: for the SPAF study, mild global or focal abnormality present; for the SPINAF study, mild global abnormality present and/or any regional wall motion categorized as hypokinetic; and for the BAATAF study, 1 or 2 regions assessed as hypokinetic or 1 region assessed as akinetic. Moderate to severe abnormality was considered present when there was a global or regional abnormality of left ventricular function worse than mild, as defined above.
Event rates were estimated based on patient-years of exposure to first ischemic stroke, death, or end of study, and analyzed by an intention-to-treat paradigm. A Cox proportional hazards model was used to assess both the univariate and multivariate association of clinical and echocardiographic parameters with the risk of stroke. In all analyses, the estimate for the hazard ratio is given as an estimate of the relative risk (RR). To account for systematic differences in measurement of the left atrium and left ventricular mass among the 3 studies, models were constructed using percentile rankings for left atrial size grouped by sex and study. A 2-tailed P≤.05 was considered statistically significant.
Among 1066 patients with AF in the pooled analysis, 78% were men, the mean age was 67 years, 55% had a history of hypertension, 19% had a history of diabetes mellitus, 7% had a previous TIA or stroke, and 27% had a history of congestive heart failure (Table 1). Moderate to severe left ventricular systolic dysfunction was present in 129 patients (13%), including 50 patients with no history of clinical heart failure.
Among these 1066 patients, 78 ischemic strokes occurred during an average of 1.6 patient-years of follow-up (annual rate, 4.7%). The only echocardiographic feature that had a strong statistically significant univariate association with stroke was moderate to severe left ventricular dysfunction (P<.001; Table 2 and Figure 1); a finding that was consistent in all 3 studies. Interventricular septal thickness also reached conventional statistical significance (P=.02). The association between increasing left atrial diameter and the risk of stroke was not statistically significant (RR, 1.02/mm; 95% confidence interval [CI], 0.99-1.06; P=.10) if left atrial size was analyzed as a continuous variable, or if the upper 25th percentiles (RR, 1.49; 95% CI, 0.90-2.48; P=.12) or 10th percentiles (RR, 0.72; 95% CI, 0.29-1.79; P=.48) were analyzed.
Independent clinical predictors of ischemic stroke in these patients were age (RR, 1.5 per decade; 95% CI, 1.1-1.9; P=.008), previous stroke or TIA (RR, 3.4; 95% CI, 1.8-6.6; P<.001), history of diabetes (RR, 1.7; 95% CI, 1.0-2.8; P=.05), and history of heart failure (RR, 1.7; 95% CI, 1.1-2.7; P=.03) (Table 3). History of hypertension, which was previously shown to predict stroke,1 was not significantly associated with stroke in the current analysis (RR, 1.5; 95% CI, 0.9-2.5; P=.11). After adjustment for these 5 clinical parameters, moderate to severe left ventricular dysfunction remained associated with an important increase in the risk of stroke (RR, 2.5; 95% CI, 1.5-4.4; P<.001) (Table 3). When adjusted for clinical predictors, left atrial diameter had no predictive value for stroke (P=.62).
A clinical risk stratification scheme previously developed by our group found that patients younger than 65 years who had no prior TIA or stroke and no history of hypertension or diabetes had an annual rate of stroke of 1%.1 Echocardiography performed in 163 such patients, 2 of whom had a stroke (annual stroke rate, 0.8%; 95% CI, 0.2%-3.0%) (Table 4), found 10 with moderate to severe left ventricular dysfunction. Their stroke rate was 9.3% per year (95% CI, 1.3%-66%) compared with 0.4% per year (95% CI, 0.06%-2.9%) in the 153 patients who were considered low risk according to clinical features and who had normal or mildly abnormal left ventricular function on echocardiography (Table 4). Conversely, 728 of the 847 patients at high risk of stroke according to clinical criteria had normal or mildly abnormal left ventricular function. Their annual risk of stroke was 4.4% (95% CI, 3.4%-5.8%). When clinical heart failure was added to this risk stratification scheme, none of the 143 patients younger than 65 years with no clinical risk factors had a stroke, including the 4 patients with moderate to severe left ventricular systolic dysfunction shown on echocardiography.
Left ventricular systolic dysfunction shown via 2-D transthoracic echocardiography was a powerful, consistent, and independent predictor of stroke in patients with AF. By contrast, neither left atrial diameter nor mitral regurgitation were significantly associated with risk of stroke in this analysis. Since our results were derived from echocardiograms obtained and interpreted by many cardiologists at participating sites, these results are likely generalizable to clinical practice. In a previous analysis7 of the SPAF I placebo-assigned patients included in this analysis, left atrial diameter was significantly associated with the combined end point of ischemic stroke and systemic emboli. However, it was not predictive of stroke in a second, larger cohort of patients with AF given aspirin in the SPAF study.8 In both of these analyses, left ventricular dysfunction emerged as a powerful independent risk factor.7,8 With our method of combining and analyzing data, left atrial diameter did not add useful independent information about the risk of stroke in patients with AF.
The clinical factors that were independently associated with stroke in this analysis of patients in 3 trials with full echocardiographic evaluations (increasing age, previous TIA or stroke, history of diabetes, or history of heart failure) were slightly different from those identified in our previous analysis from all 5 primary prevention trials (increasing age, previous TIA or stroke, history of diabetes, or history of hypertension).1 In both analyses, history of hypertension and history of heart failure were associated with risk of stroke by univariate analysis. Thus, the difference in clinical predictors in the 2 analyses may be more apparent than real. This highlights both the problem of relying excessively on an arbitrary P value as evidence of an association, and the relatively small sample size in this analysis. For consistency, the originally published clinical risk stratification scheme derived from all 5 trials1 was used to assess the additional information provided by echocardiography (Table 4).
The data in Table 4 must be interpreted with caution. The small number of patients in many of the risk categories may lead to unstable estimates of the risk of stroke. For example, it is unlikely that patients older than 75 years who have no clinical risk factors and normal left ventricular systolic function truly have a lower rate of stroke than younger patients with similar risk profiles. Nevertheless, normal left ventricular function generally decreased the risk of stroke, and moderate to severe left ventricular dysfunction increased this risk. Clinicians may decide to administer anticoagulants to some patients with poor left ventricular systolic function whom they would otherwise have considered to be at low risk of stroke. Similarly, knowledge of left ventricular function is useful when weighing the risks and benefits of anticoagulation in patients who have a high risk of bleeding.
In summary, although the results of transthoracic echocardiography are unlikely to change the decision to administer anticoagulants for most patients, knowledge of left ventricular systolic function may be useful in a small number of patients.
Accepted for publication November 20, 1997.
We thank all the investigators and the Statistics and Epidemiology Research Corp, Seattle, Wash; John Marler, MD, and Michael Walker, MD, of the National Institute of Neurological Disorders and Stroke; and Ina Sungaila of DuPont-Merck Pharmaceuticals, Mississauga, Ontario.
Writing Committee: Michael Ezekowitz, MD; Andreas Laupacis, MD; Gudrun Boysen, MD; Stuart Connolly, MD; Robert Hart, MD; Kenneth James, PhD; Philip Kistler, MD; Peter Koudstaal, MD; Richard Kronmal, PhD; Ruth McBride; Palle Petersen, MD; Daniel Singer, MD.
Echocardiographic Committee: Richard Asinger, MD (SPAF); Joseph Blackshear, MD (SPAF); Michael Ezekowitz, MD, PhD (SPINAF); Jonathan Plehn, MD (BAATAF).
Atrial Fibrillation, Aspirin, Anticoagulation Study: Palle Petersen, MD; Gudrun Boysen, MD; John Godtfredsen, MD; Ellen Andersen, MD; Bjorn Andersen, MD.
Stroke Prevention in Atrial Fibrillation Study (SPAF Investigators)
Boston Area Anticoagulation Trial for Atrial Fibrillation Study (BAATAF): Daniel Singer, MD; Robert Hughes, MD; Daryl Gress, MD; Mary Sheehan, RN; Lynn Oertel, RN-C; Sue Maraventano, RN; Dyan Blewett, MS; Bernard Rosner, PhD; Philip Kistler, MD.
Canadian Atrial Fibrillation Anticoagulation Study: Stuart Connolly, MD; Andreas Laupacis, MD; Michael Gent, DSc; R. Roberts, MSc; J. Cairns, MD; C. Joyner, MD; and the CAFA Investigators.
Stroke Prevention in Nonrheumatic Atrial Fibrillation Study: Michael Ezekowitz, MD; Samuel Bridgers, MD; Kenneth James, PhD; Cindy Colling, RPH; Heidi Krause-Steinrauf, MS; and the SPINAF Investigators.
The European Atrial Fibrillation Study (EAFT): Anet van Latum; Peter Koudstaal; and the EAFT Investigators.
Reprints: Michael D. Ezekowitz, MD, PhD, Cardiovascular Medicine, Yale University, 333 Cedar St, 3 FMP, PO Box 3333, New Haven, CT 06510-8056.