Atrial fibrillation after cardiac surgery

A frequent complication of cardiac surgery is postoperative atrial fibrillation, which often results in morbidity, prolonged hospitalization, and death.^1,2 Postoperative atrial fibrillation is estimated to cost more than $1 billion in the United States annually.² It has been shown to predict subsequent stroke, congestive heart failure, and late atrial fibrillation among patients undergoing mitral valve surgery.³ One study showed that postoperative atrial fibrillation occurred in 62% of patients after combined aortic valve and coronary artery bypass graft (CABG) surgery, in 49% of patients after aortic valve surgery alone, in 42% of patients after mitral valve surgery, and in 32% of patients after CABG surgery alone. Most risk indexes for the prediction of postoperative atrial fibrillation include several intraoperative and postoperative factors to achieve sufficient discriminative power.^4,5 Such algorithms do not adequately risk stratify patients or allow for preventive preoperative treatment.

Measurement of left atrial volume has been shown to portend the development of atrial fibrillation in nonsurgical situations,⁶ and it is a measurement that can be obtained without difficulty before surgery. An increase in atrial volume is an indicator of chronically elevated left ventricular filling pressure,⁷ which reflects deterioration of cardiovascular diastolic function. We conducted a study to determine whether left atrial volume, as a measure of the immediate increase in filling pressure before cardiac surgery, would predict atrial fibrillation after cardiac surgery.

Subjects and methods

P

P

t

We prospectively enrolled consecutive patients scheduled for myectomy, CABG surgery, and surgery involving the aortic valve, ascending aorta, tricuspid valve, pericardium, and cardiac masses. Those with previous or scheduled Cox’s maze procedure, mitral valve prosthesis, more than moderate mitral valve stenosis or regurgitation, congenital cardiac abnormalities, and chronic atrial fibrillation were excluded from the study. Using conventional techniques, preoperative echocardiograms were examined in a blinded manner.^8,9 To measure left atrial volume, we used the biplanar area-length technique and indexed to body surface area.⁹ Subjects with a left atrial volume > 32 mL/m², which is the upper 95th percentile of normal subjects, were considered to have left atrial enlargement. Using biplanar Simpson’s method, left ventricular ejection fraction was determined from the same views. To assess quality assurance, 10% of subjects were randomly chosen to be reevaluated by one of the investigators. The linear correlation (r = 0.87; < .001) was good, and the mean difference of —1.03 mL/m² was not different from zero ( = .47). Doppler evaluation of cardiac hemodynamics was focused on cardiac diastolic function. We assigned 5 diastolic function grades (DFGs) based on these measurements.⁸ All atrial fibrillation events were confirmed by either 12-lead electrocardiogram or telemetry recordings. If an episode lasted longer than 30 seconds, it was considered clinically relevant. Comparisons between groups for continuous variables were performed using Student’s test and Mann Whitney-U test. Cox proportional hazards models and the Kaplan-Meier product limit method were used to model the main outcome of postoperative atrial fibrillation. Sensitivity, specificity, positive predictive value, and negative predictive value of emerging dichotomous predictors of postoperative atrial fibrillation were calculated. Receiver-operator characteristics (ROCs) were generated for continuous and categorical variables.

Results

The sample consisted of 205 subjects, 34.6% of whom were women. The mean age was 62 ± 16.3 years. The

P

shows the baseline characteristics and their association with postoperative atrial fibrillation in Cox regression analyses. Subjects with postoperative atrial fibrillation were more likely to be older and to have a history of chronic obstructive pulmonary disease, a greater degree of left main coronary artery stenosis, and higher systolic and diastolic blood pressure. Among the 203 subjects available for follow-up after surgery, 84 developed postoperative atrial fibrillation after a median of 1.8 days. Thirty-seven subjects had persistent atrial fibrillation lasting longer than 24 hours, 14 had multiple episodes, and 33 had a single episode. Subjects with postoperative atrial fibrillation had a median hospital stay of 7 days, compared with 5 days for those without postoperative atrial fibrillation ( < .001). Left atrial volume and DFG were the only measurements among all echocardiographic parameters that predicted postoperative atrial fibrillation. DFG had the strongest predictive value when considered as normal vs abnormal (hazard ratio [HR] = 3.36). Left atrial volume > 32 mL/m² was a very strong predictor of atrial fibrillation (HR = 6.55). As shown in

P

, with each quartile of left atrial volume, the percentage of subjects with any type of postoperative atrial fibrillation increased, as did the subgroup of persistent atrial fibrillation. Apart from age, left atrial volume indexed to body surface area was the only independent predictor of postoperative atrial fibrillation, as shown on a multivariable Cox proportional hazards model. Subjects with an abnormal left atrial volume (> 32 mL/m²) had almost 5 times the risk of developing postoperative atrial fibrillation, even following adjustment for other risk factors (adjusted HR = 4.84; 95% confidence interval, 1.93-12.17; = .001). When only persistent postoperative atrial fibrillation was considered, age and left atrial volume were equally strong as independent predictors of postoperative atrial fibrillation. Therefore, we stratified the population to obtain an optimal separation between subjects with and without postoperative atrial fibrillation (

P

). Among those with normal left atrial volume, only 5 subjects older than 65 years and no subject younger than 65 years had postoperative atrial fibrillation. Almost all atrial fibrillation events occurred in older subjects with enlarged atria. In an ROC analysis, the area under the curve for left atrial volume alone was 0.729, and for the combination of age and the 4 strata of left atrial volume ordered as shown in Figure 1 was 0.768 (both < .001). The sensitivity for identifying subjects with postoperative atrial fibrillation was 76.2%, and the specificity was 71.4%, with a positive predictive value of 65.5% for the combined presence of age older than 65 years and left atrial volume > 32 mL/m². The negative predictive value of age older than 65 years and left atrial volume > 32 mL/m² was 81.0%.

TableFigure 1Figure 2

Discussion

In this study, the strongest risk factor for postoperative atrial fibrillation was left atrial volume indexed to body surface area. This association was independent of age and other surgical or clinical variables. After adjusting for other risk factors, subjects with an abnormal left atrial volume had almost 5 times the risk of developing postoperative atrial fibrillation. The combination of age older than 65 years and left atrial volume < 32 mL/m² effectively identified subjects at risk for postoperative atrial fibrillation.

Recently, several researchers have suggested various risk-scoring strategies to predict the development of postoperative atrial fibrillation.^4,5 Although some have achieved a reasonable measure of success, they used an involved point-scoring system, with up to 8 variables.⁴ In these systems, mostly intraoperative or postoperative factors were shown to be the strongest predictors of postoperative atrial fibrillation. In addition, the generalizability of a number of the studies was limited to certain forms of surgery³ or by the exclusion of patients with a history of atrial fibrillation,^3,5,10 previous cardiac surgery,^3,5 or antiarrhythmic therapy.¹⁰ The use of these studies clinically is limited, although they do provide some insight into the development of postoperative atrial fibrillation.

Diastolic dysfunction was a significant univariate predictor of postoperative atrial fibrillation in our study. A positive trend was also shown for early diastolic myocardial relaxation velocity E', which has been demonstrated to correlate with left ventricular filling pressure. Plasma B-type natriuretic peptide (BNP) level, another indicator of ventricular filling pressure, was shown to moderately predict postoperative atrial fibrillation.¹⁰ These indicators, however, are strongly dependent on loading conditions and are only an indication of a temporary state. B-type natriuretic peptide is particularly variable when a patient is undergoing major surgery. Therefore, BNP levels and diastolic function are not long-term, stable characteristics that can indicate risk under varying conditions, although they may provide a valid indication of the patient’s pathophysiologic state before surgery. Thus, the incremental value of acutely measured filling pressure at a certain time before surgery and its potential to provide clinicians with useful cut-offs for risk stratification beyond clinical risk factors has to be questioned.

The onset of nonsurgery-related atrial fibrillation has been shown to be preceded by a measurable increase in left atrial volume, as demonstrated in our and other studies.⁶ Enlarged atria are best correlated with increased wall tension due to intermittent yet chronic elevation of ventricular filling pressures.⁷ The correlation between diastolic dysfunction and left atrial volume as an indicator of the severity and duration of disease has been likened to the relationship between glycosylated hemoglobin and serum glucose.¹¹ Fibrosis, collagen production, and the intercellular matrix are increased by chronic myocyte stretch, which is mediated through the renin-angiotensin-aldosterone system.¹² Increased atrial volume is indicative of this remodeling process and is a sign of a permanent arrhythmogenic substrate. These changes can be anticipated to make the cardiovascular system more vulnerable to an increased adrenergic state and large volume changes, such as those that occur during open heart surgery.

As in nearly all other studies on postoperative atrial fibrillation, age was the only other risk factor apart from left atrial volume that was shown to be significant in our study.^3-5 The area under the curve was 0.77 for risk prediction based on the 2 strictly preoperative factors of left atrial volume and age. This level of discriminative power had only been obtained previously in a study that used an involved strategy of scoring predominantly postoperative factors.⁴

The use of beta blockers or amiodarone (Cordarone), or both, is currently recommended for the prevention of postoperative atrial fibrillation. Amiodarone has been investigated as a more aggressive alternative option to target postoperative atrial fibrillation. In the Prophylactic Oral Amiodarone for the Prevention of Arrhythmias That Begin Early After Revascularization, Valve Replacement, or Repair (PAPABEAR) trial, amiodarone was shown to significantly decrease the incidence of postoperative atrial fibrillation.¹³ No change was observed, however, between the placebo and amiodarone groups regarding the duration of hospital stay, the longest atrial fibrillation episode, or the time of onset of atrial fibrillation. The number of drug-related adverse events was significantly higher in the amiodarone group.

From these intervention trials, it becomes clear that the current approach to postoperative atrial fibrillation is unsatisfactory. Prevention does not appear to be directed at the patient population at risk and is not tailored to address the pathophysiologic basis for postoperative atrial fibrillation that characterizes this subset of cardiac surgery patients. These observations and those noted in our study support the contention that there is a more pervasive physiological abnormality, which accounts for cardiovascular events, including atrial fibrillation after cardiac surgery. Reduction of increased left ventricular filling pressures and prevention of atrial remodeling are promising approaches to reduce the incidence of postoperative atrial fibrillation. Angiotensin-converting enzyme inhibitors¹⁴ and angiotensin II receptor blockers¹⁵ are known to reduce the incidence of atrial fibrillation in nonsurgical populations and are worthy of investigation in the context of postoperative atrial fibrillation.

Conclusions

As a marker of chronic diastolic dysfunction, left atrial volume is strongly and independently related to the occurrence of atrial fibrillation after cardiac surgery. A simple algorithm based on left atrial volume and age allows for effective risk stratification before surgery to target the patient population that would benefit from preventive therapy.

Acknowledgments

Dr Osranek was supported by postdoctoral fellowships from the Austrian Science Fund (Schrödinger Stipend) and the American Heart Association.