Methods

Consecutive patients with nonvalvular persistent AF, submitted to successful electrical cardioversion at our department between January 2008 and January 2009, were considered for inclusion in this study. According to American College of Cardiology/American Heart Association/European Society of Cardiology guidelines persistent AF was defined as a nonself-terminating episode (>7 days), and successful cardioversion was identified by restoration of stable sinus rhythm with no evidence of immediate AF recurrence within the first 2 minutes after cardioversion. Exclusion criteria were cardioversion of recurrent AF (defined as ≥1 documented previous episode of persistent AF), overt coronary artery disease (defined as ≥1 of the following: history of acute coronary syndrome or coronary revascularization, segmental wall abnormalities, or positive exercise stress test result), severe LV systolic dysfunction (defined as ejection fraction ≤30%), atrial flutter, valvular stenosis of any degree, moderate or severe valve regurgitation, valvular prosthesis, AF with identifiable reversible causes, high-degree atrioventricular block, pacemaker implantation, thyroid disorders, and patient refusal to participate in the study. The study was conducted in accordance with the Declaration of Helsinki and informed consent was obtained by all patients.

All patients underwent transthoracic echocardiography immediately before electrical cardioversion using a Vivid 7.0 (GE Healthcare, Horten, Norway) ultrasound system. LV dimensions and thicknesses at end-diastole and end-systole were measured in accordance with current American Society of Echocardiography guidelines. LV volumes and ejection fraction were calculated from apical 4- and 2-chamber views using the modified Simpson rule. Left atrial volume was calculated from apical 4- and 2-chamber views using the biplane method of disks. LV and left atrial volumes were indexed to body surface area. Standard pulsed Doppler interrogation of LV inflow was performed from the apical 4-chamber view by placing the sample volume at the level of the mitral leaflet tips. E-wave deceleration time was measured and used as a standard index of LV diastolic function. Pulmonary artery systolic pressure was estimated by measuring peak tricuspid regurgitation velocity and estimating right atrial pressure in accordance with inferior vena cava size and respiratory collapse. Systolic right ventricular function was explored by measuring tricuspid annulus plane systolic excursion.

Pulsed tissue Doppler imaging of mitral annulus motion was performed from the apical 4-chamber view by placing a 5-mm sample volume at the junction between the basal lateral myocardium and the adjacent annulus. Filter and gain settings were adjusted at the minimal level for optimum signal-to-noise ratio. Peak systolic mitral annulus velocity and E _m were measured. E _m was used as a relatively preload-independent measurement of LV relaxation, and E/E _m was assumed as an index of LV filling pressure. An E/E _m ≤8, which is recommended by current American Society of Echocardiography guidelines as a cutoff to identify patients with normal LV filling pressure, was used to dichotomize patients and to plot Kaplan–Meier curves describing probability of sinus rhythm maintenance over time. All measurements were calculated by averaging data obtained in 5 consecutive beats. To minimize risk of angle-dependent errors particular care was given to checking for adequate alignment of mitral annulus motion with the ultrasonic beam.

Reproducibility of tissue Doppler measurements in our laboratory was previously reported. For this study variability was assessed in a randomly selected subset of 15 subjects. Intraobserver coefficients of variation were 3.3% for E _m and 3.7% for E/E _m . Corresponding intraclass correlation coefficients were 0.98 and 0.97, respectively (p <0.0001 for the 2 comparisons). Interobserver coefficients of variation were 6.7% for E _m and 7.3% for E/E _m , with corresponding intraclass correlation coefficients of 0.94 and 0.93, respectively (p <0.0001 for the 2 comparisons).

After successful sinus rhythm restoration, remote electrocardiographic monitoring was performed for ≥24 hours to document maintenance of sinus rhythm. In accordance with current guidelines, the decision to prescribe antiarrhythmic prophylaxis was taken after an accurate global evaluation of risk of AF relapse by an investigator blinded to results of tissue Doppler assessment. Treatment with nonantiarrhythmic cardiovascular agents was kept unchanged in all patients. All patients received appropriate oral anticoagulation ≥3 weeks before and 4 weeks after electrical cardioversion (target international normalized ratio 2.0 to 3.0).

All patients were then followed 2 times a week for the first month and then had monthly visitations until 1 year after cardioversion. A targeted clinical examination and a 12-lead electrocardiogram were performed at each visit. A 24-hour ambulatory electrocardiographic monitoring was performed in all patients who were in sinus rhythm at 3-month follow-up. Predefined end points of this study were recurrent persistent AF at 2-week follow-up (early recurrence of AF [ERAF]) and at 1-year follow-up (i.e., including ERAF and late recurrence of AF [LRAF]).

Data are reported as mean ± SD for normal variables and as median (interquartile range) for nonparametric variables. For time-independent analysis logistic regression analysis was performed using the maximized log likelihood method to evaluate the independent effect of tissue Doppler diastolic indexes on risk of AF recurrence. Diagnostic colinearity with analysis of tolerance and variance inflation factor was performed to assess model stability. Risk of AF relapse was expressed as a function of tissue Doppler indexes using the formula p=1−1/(e[a+b×X]+1)p=1−1/(e[a+b×X]+1)
p = 1 − 1 / ( e [ a + b × X ] + 1 )
, where p is the estimated probability of AF recurrence, e is the natural logarithmic base, a and b are the intercept and slope of the logistic regression line, respectively, and X indicates the tissue Doppler measurement of LV diastolic function. Values corresponding to estimated 25%, 50%, and 75% probabilities of recurrence were determined using the formula (logit[p]−a)/b
( logit [ p ] − a ) / b
, where logit (p) is the log odds ratio (OR), calculated as the natural logarithm of (p/[1 – p]). To estimate risk changes for any unitary variation in tissue Doppler indexes, a linear approximation was assumed for curves in the interval corresponding to 25% and 75% risks of AF recurrence. Linearity in this range was assessed by fitting data using linear regression analysis based on the least squares method. Logarithmic transformation followed by assessment of normality using Shapiro–Wilks test was performed for nonparametric variables before insertion in regression analyses.

For time-dependent analysis the independent effect of variables on risk of cardiovascular events was explored by univariate Cox regression. For multivariate Cox regression the following procedure was used. All variables that showed a p value <0.10 at univariate Cox analysis were tested in a multivariate stepwise analysis. The stepwise selection method was based on score statistics for entry testing and on Wald statistic for removal. A score entry statistic ≤0.05 and a Wald removal statistic ≥0.10 were used as selection criteria during the stepwise procedure. Goodness-of-fit of models was assessed by exploring the distribution of Cox–Snell residuals, whereas partial residuals were plotted against time to test the proportional hazard assumption. Differences in cumulative event-free survival probability between patients with normal and those with an increased E/E _m were explored using the Kaplan–Meier method followed by log-rank test.

The sample size for this study was calculated by assuming that logistic regression would have identified a significant association between tissue Doppler diastolic indexes and risk of LRAF. Hypothesizing an expected 50% overall prevalence of 1-year AF recurrence, 40 patients would have been necessary to achieve 90% power in detecting a threefold risk increase for a 1-SD increase in the predictor variable. Assuming SDs <4 cm/s for E _m and <4 for E/E _m , this would have corresponded to ORs <0.760 for E _m and >1.316 for E/E _m .

A p value <0.05 was considered statistically significant. All tests were 2-tailed. SPSS 15.0 for Windows (SPSS, Inc., Chicago, Illinois) was used to generate statistical analyses.