Background
The clinical assessment of left atrial function during atrial fibrillation is challenging and often inaccurate because of the beat-to-beat variability in the cycle length. The aim of this study was to validate the use of an index beat, the beat following two preceding cardiac cycles of equal duration, for the measurement of left atrial functional indices, including area, volume, and expansion index. The index beat was compared with the conventional but time-consuming method of averaging multiple consecutive cardiac cycles.
Methods
Thirty patients with persistent or permanent atrial fibrillation were studied using two-dimensional echocardiography, and left atrial indices were measured from the average of 17 consecutive cardiac cycles compared with that of an index beat taken from outside of these 17 cycles.
Results
The index beat showed good correlation with the averaging technique, and comparison of the two methods showed them to be interchangeable. Clinically, the differences in left atrial functional indices between the two methods were minor.
Conclusions
Use of the index beat to measure dynamic left atrial function in atrial fibrillation can easily be performed and is as accurate as and less time consuming than the onerous method of averaging of multiple cardiac cycles.
The assessment of left atrial (LA) size and function in the setting of atrial fibrillation (AF) is difficult because of the loss of atrial contraction and also the beat-to-beat changes in preload, afterload, and left ventricular (LV) contractility that invariably affect atrial diastolic filling and systolic ejection into the ventricle. Despite this, the assessment of LA function in AF provides important information about the extent of the diseased atrial substrate, guides therapy, and helps predict outcome after interventions to restore sinus rhythm. Therefore, methods to accurately and simply quantify LA size and function are still being sought.
Traditionally, for the accurate and reproducible assessment of both atrial and ventricular function in AF, the widely accepted method has been to take the average of measurements taken over multiple consecutive cardiac cycles. It has been suggested that the optimal number of beats required to estimate LV systolic function in a reproducible manner in AF is ≥13 beats for <5% variability in the measured values, or roughly three times that necessary in sinus rhythm. However this method is not only tedious but variable, dependent on the window of cardiac cycles selected (i.e., windows with short RR intervals may inaccurately represent atrial filling and emptying capacity).
To overcome this onerous technique of averaging multiple cardiac cycles for each LA measurement, we propose the use of an “index beat,” the beat following a pair of equal preceding cardiac cycles (RR1 and RR2), where RR1/RR2 = 1 or approximately 1, to substitute the averaging method in estimating stable atrial function in patients with AF.
The use of an index beat has been shown to reliably and reproducibly estimate LV systolic function in AF despite the varying influence of preload and afterload. These reports show that LV repolarization and restitution is complete only after 500 msec, and cycle lengths shorter than this do not accurately represent stable ventricular function in a reproducible manner. To date, validation of the index beat has not been performed in the assessment of LA function in patients with AF. The purpose of this study was to estimate LA function using an index beat and correlate these with the average of measurements taken over multiple cardiac cycles, in the hope of providing a simplified yet accurate alternative for LA functional assessment using two-dimensional (2D) echocardiography.
Methods
Study Population
Thirty patients (mean age, 64 ± 12 years; 22 men) with persistent or permanent nonvalvular AF referred for 2D echocardiography were prospectively evaluated. Exclusion criteria included poor acoustic windows, inability to hold the breath in full expiration for 15 to 20 sec, and a resting ventricular rate > 90 beats/min during AF. Patients were further subdivided into groups on the basis of the CHADS 2 stroke risk score (congestive cardiac failure, hypertension, age > 75 years, diabetes, and stroke history). Each stroke risk factor acquires 1 point, except stroke history, which gains 2 points, for a maximum score of 6 points. A CHADS 2 score > 2 carries a significant risk for stroke, vascular events, and heart failure, with unequivocal need for long-term anticoagulation. Patients were divided by CHADS 2 score < 2 or ≥ 2. Baseline characteristics of all study patients are presented in Table 1 . All patients provided informed consent, and the study was approved by the local research and ethics committee.
| Variable | Value |
|---|---|
| Age (yrs) | 64 ± 12 |
| Men | 73% |
| Height (cm) | 173 ± 11 |
| Weight (kg) | 85 ± 18 |
| BMI (kg/m 2 ) | 29 ± 5 |
| AF duration (mo) | 34 ± 18 |
| CHADS 2 score | 1.43 ± 0.85 |
| Heart rate (beats/min) | 79 ± 8 |
| Hypertension | 22 (73%) |
| Diabetes mellitus | 9 (30%) |
| Coronary artery disease | 13 (43%) |
| LA diameter (long) (cm) | 6.2 ± 0.7 |
| LVEF (%) | 60 ± 7% |
| β-blockers | 17 (56%) |
| Digoxin | 9 (30%) |
| Calcium channel blockers | 10 (33%) |
Echocardiography
All enrolled subjects underwent comprehensive 2D transthoracic echocardiography using a Vivid 7 ultrasound scanner (GE Vingmed Ultrasound AS, Horten, Norway). An experienced cardiologist performed all echocardiographic studies. For each patient, two sets of 17 consecutive cardiac cycles for each atrial view were recorded with breath held at full expiration. An electrocardiogram was recorded simultaneously at a sweep speed of 100 mm/sec. All measurements were performed offline by two cardiologists using an EchoPAC 6.1 workstation (GE Vingmed Ultrasound AS). Each observer was not permitted to watch the other perform the measurements on the collected data, to remove any possible influence on the other observer.
LA diameter was obtained from the apical four-chamber view in the long and short dimensions after the offset of the T wave from the surface electrocardiographic recording, according to American Society of Echocardiography guidelines. Measurement of LA area and volumes were performed in the apical four-chamber and two-chamber views by manually outlining the atrial endocardial cavity ( Figure 1 ). The average of the 17 cycles from the four-chamber and then two-chamber view were collected from each patient, and the mean of these two values produced the biplane values for area and volumes. When the pulmonary veins or LA appendage was visualized, these were excluded from the outline. Values were indexed to body surface area. Minimum atrial volume was measured from just after the onset of the QRS complex, immediately after mitral valve closure. Maximum atrial volume was measured at the offset of the T wave immediately before mitral valve opening for each cardiac cycle. Because atrial contraction was absent in all patients, atrial ejection fraction could not be calculated, and because reservoir function is a significant contributor to LA phasic function in AF, atrial filling was instead estimated by use of the atrial expansion index. The LA expansion index was calculated as [(LA maximum volume − LA minimum volume)/LA minimum volume] × 100.
All atrial parameters were measured from the left atria over 17 consecutive cardiac cycles free of ventricular extrasystolic beats, and the average of these was calculated.
Index Beat Selection
The index beat taken following two equal preceding cardiac cycles was selected from outside of the 17 consecutive cardiac cycles. Pairs of equal cardiac cycles >500 msec in duration were selected as index beats, allowing for a maximal difference of 60 msec between preceding (RR1) and pre-preceding (RR2) cardiac cycles ( Figure 1 ).
Reproducibility of Atrial Expansion Index Measurements
All 30 studies were selected for analysis by two independent cardiologists on two separate occasions. Both observers repeated all measurements on the 30 patients 1 week apart to determine the degrees of interobserver and intraobserver variability for atrial indices.
Statistical Analysis
Continuous variables are presented as mean ± SD. Differences in continuous variables between the averaging method and the index beat measurements and between subgroups of CHADS 2 scores were assessed using independent-samples t tests. Categorical variables are presented as absolute numbers and percentages, and the statistical test used was the χ 2 test.
To assess the variability and accuracy of the mean of 17 consecutive cardiac cycles, we calculated the standard error of the mean (SEM) of 17 cardiac cycles for each parameter. Analysis of variance was used to calculate the intrasubject SD of the cardiac cycles for each atrial parameter. The SEM was then calculated as (within-subjects SD)/√(number of cardiac cycles), where the number of cardiac cycles was 17.
The variability of each atrial parameter over 17 cycles was expressed as a coefficient of variation (i.e., percentage of the SE of the average over the grand mean).
For comparison of agreement in measurements by the two different methods (multiple cardiac cycles vs index beat), Bland-Altman analysis was applied to evaluate systematic differences and limits of agreement between atrial parameters. For each parameter, the 17-cycle average of each observer was taken and then averaged and compared with the average of the index beat measurements taken from both observers. Linear regression analysis was used to assess the relationship between the index beat with the average of multiple cardiac cycles and also to assess if any bias in the Bland-Altman analysis was independent of the measurements themselves.
Furthermore, the effect of varying RR intervals on the congruence of the two methods (averaging vs index beat) was assessed by comparing results on the basis of high and low SDs of the 17 RR intervals. Subjects were divided into two groups on the basis of SD variation. Group 1 included subjects with low SD variation (<100 msec; n = 17 patients), and group 2 had SD variation >100 msec across 17 RR intervals ( n = 13 patients). Bland-Altman analysis was then performed on biplane LA volume (maximum and minimum), comparing the two methods within the two groups to demonstrate differences in limits of agreement on the basis of SD variation over 17 RR intervals.
Bland-Altman plots were also constructed for interobserver and intraobserver variability to obtain 95% upper and lower limits of agreement, bias, and 95% confidence intervals of bias. Because this method calculates the SD of the difference in any two given measurements and includes the bias, a repeatability coefficient was not required. P values < .05 were considered statistically significant. Statistical analysis was performed using SPSS version 18.0 (SPSS, Inc., Chicago, IL) and Analyse-it version 2.22 (Analyse-it Software, Ltd., Leeds, United Kingdom) for Bland-Altman plots.
Results
All 30 patients studied had adequate images for four-chamber analysis. Only 24 patients had complete 17-cycle data sets for two-chamber analysis. The mean age was 64 ± 12 years, and the cohort was predominantly male (73%). The average duration of AF was 34 ± 18 months. Baseline characteristics are presented in Table 1 .
Repeatability of the 17 cardiac cycles showed high accuracy, with variability (SEM over grand mean coefficient of variation) ranging from 2% to 8% for all individual LA parameters ( Table 2 ). Because the expansion index had to be calculated from the 17-cycle average of two variables, namely, minimum and maximum LA volumes, the variability increased to 13%.
| Variable | Mean ± within-patient SD | SEM | Coefficient of variation (SEM/mean) |
|---|---|---|---|
| LAD short (cm) | 4.9 ± 0.5 | 0.1 | 2.5% |
| LAD long (cm) | 6.2 ± 0.7 | 0.2 | 2.8% |
| LAA (cm 2 ) | 27 ± 6 | 1.4 | 5.2% |
| LAV max (mL) | 96 ± 30 | 7 | 7.5% |
| LAV min (mL) | 77 ± 25 | 7 | 7.9% |
| LAEI (%) | 29 ± 15 | 4 | 12.8% |
Dynamic LA function as assessed by the index beat, where the equal RR1 and RR2 cardiac cycles were >500 msec, gave similar absolute measurements with the average of 17 cardiac cycles ( Figure 2 ). Nor were there any differences when the study cohort was divided on the basis of CHADS 2 score ≥ 2 or < 2, which served to distinguish clinical risk for more diseased atrial substrates ( Table 3 ).
| LA measure | Group 1 | Group 2 | P |
|---|---|---|---|
| LAV max | |||
| Average of 17 cycles | 103 ± 32 | 90 ± 31 | .31 |
| Index beat | 103 ± 32 | 90 ± 31 | .31 |
| LAV min | |||
| Average of 17 cycles | 82 ± 30 | 72 ± 31 | .33 |
| Index beat | 83 ± 30 | 72 ± 31 | .32 |
| LAEI | |||
| Average of 17 cycles | 25 ± 26 | 34 ± 13 | .25 |
| Index beat | 25 ± 25 | 33 ± 13 | .31 |
No significant systematic bias was detected between the two methods, and any small bias was negligible, because the confidence intervals included zero and the limits of agreement were relatively narrow ( Figure 3 ). In Figure 3 , the Bland-Altman plots show that the data points are close to the bias line for lower values as for higher values; therefore, there appears to be no increase in bias or error at higher or lower data points with either method for all measured parameters. Accordingly, logarithmic transformation was not performed.
The linear regression slope did not significantly deviate from 1 in most measured parameters, suggesting that a unit of change in one method corresponds to a similar unit of change with the second method (i.e., if the index beat method gives higher values for a given data point, the averaging of multiple cardiac cycles also gives higher values to the same extent). Therefore, also considering that biases were small, the index beat approach does not appear to overestimate or underestimate the averaging method.
Furthermore, the effect of RR interval variation on the accuracy of index beat assessment was evaluated. The SD of the RR intervals ranged from 59.14 to 153.85 msec. The range of SDs across 17 RR intervals in group 1 was 59.14 to 99.70 msec and in group 2 was 105.76 to 153.85 msec.
The 95% limits of agreement for maximum LA volume in the group with SDs < 100 msec comparing the two methods was −0.56 to +0.56, while in the group with SDs > 100 msec, the limits of agreement were −0.39 to +0.49. For minimum LA volume, the limits of agreement in group 1 were −1.9 to +1.7 and in group 2 were −2.0 to +1.8. These results show no significant differences between the two methods in patients with low and high SD variation in RR intervals during AF.
Interobserver and Intraobserver Variability
The analysis of interobserver and intraobserver variability demonstrated high congruence ( Table 4 ). Bland-Altman plots showed no evidence of systematic differences regarding interobserver and intraobserver variability.
