Atrial Fibrillation After Isolated Coronary Surgery Affects Late Survival
Background— Atrial fibrillation (AF) after coronary artery bypass graft surgery is a difficult problem and a continuing source of morbidity and mortality. However, the prognostic implications of postoperative AF are still in dispute. Our aim was to ascertain the impact of AF after coronary artery bypass graft on postoperative survival and to assess its prognostic role in cause-specific mortality.
Methods and Results— We conducted a prospective observational study of 1832 patients undergoing isolated coronary artery bypass graft between January 2000 and December 2005 at 2 cardiac surgery centers in northern Italy. Patients affected by postoperative AF were identified and followed up until death or study end (April 30, 2007). A total of 570 patients (31%) developed AF after coronary surgery. Patients affected by postoperative AF experienced a longer hospital stay (7 days [25th to 75th percentile, 7 to 10 days] versus 7 days [25th to 75th percentile, 6 to 8 days]; P<0.001). Hospital mortality also was higher in AF patients (3.3% versus 0.5%; P<0.001). On discharge, 1806 patients were alive; 143 were lost to follow-up. The remaining 1663 were followed up for a median of 51 months (25th to 75th percentile, 41 to 63 months); 126 of them died after a median of 14 months (25th to 75th percentile, 5 to 32 months). Long-term mortality rates were significantly higher for patients with postoperative AF (2.99 per 100 person-years; 95% confidence interval, 2.33 to 3.84; 61 deaths) compared with those without the arrhythmia (1.34 per 100 person-years; 95% confidence interval, 1.05 to 1.71; 65 deaths), with an adjusted hazard ratio of 2.13 (P<0.001) and 2.56 (P=0.001) when also accounting for the prescription of warfarin at discharge. With Cox regression, patients with AF were shown to be at higher risk of dying from embolism (adjusted hazard ratio, 4.33; 95% confidence interval, 1.78 to 10.52) but not from other causes.
Conclusions— Postoperative AF affects early and late mortality after isolated coronary artery bypass graft surgery. Patients affected by AF are at higher risk of fatal embolic events. Careful postoperative surveillance with a specific antiarrhythmic and antithrombotic prophylaxis, aimed at reducing AF and its complications, is recommended.
Received March 6, 2008; accepted July 29, 2008.
Despite ongoing efforts to decrease its occurrence, atrial fibrillation (AF) after cardiac surgery remains a difficult problem and a continuing source of morbidity.1–3 Postoperative AF can be a cause of significant adverse effects and prolongs the length of stay in intensive care and in hospital, leading to a substantial disease burden and medical costs.1,3–5
Clinical Perspective p 1618
Although many studies have examined the impact of AF on the perioperative period, none has managed to present compelling data that support an independent association between postoperative AF and late mortality.2,3,6,7 Whereas the hazards resulting from chronic AF are well established, the prognostic implications of AF after cardiac surgery are less certain.8 Several factors related to postoperative AF also predict early and late mortality.2,3,6,7 Distinguishing the complications that arise as a consequence of postoperative AF from risks related to preexisting cardiac disease is a complex issue.8
The objective of this study was to ascertain the impact of postoperative AF on postoperative survival and to asses its prognostic role on cause-specific mortality.
End Points and Study Design
This was an observational, prospective, 2-center follow-up study aimed at ascertaining the impact of AF after coronary artery bypass graft surgery (CABG) on postoperative survival and assessing its prognostic role on cause-specific mortality.
Between January 2000 and December 2005, we enrolled 1878 consecutive patients undergoing isolated CABG at 2 cardiac surgery centers in northern Italy. Forty-six patients had a history of permanent AF and were excluded from the analysis, leaving a study population of 1832 patients.
All data were prospectively collected and recorded onto computerized database registries that remained consistent over the study period. The registries captured all of the cardiac surgery that had been performed at both institutions and were validated annually by random audit of patient charts. Demographic characteristics, clinical details, and perioperative variables were supplemented by review charts, operative notes, and ECG and echocardiographic reports.
The study protocol was in compliance with the Institutional Review Board at each institution and received full approval. Oral informed consent was obtained from all patients.
The authors had full access to and take full responsibility for the integrity of the data. All authors have read and agree to the manuscript as written.
Patient management and anesthetic and surgical techniques were standardized for all patients as previously described.9 Preoperative statin therapy was not routinely administered in either center. In brief, all patients were monitored daily until discharge with continuous ECG telemetry and standard 12-lead ECG. Additional recordings were collected at clinical suspicion of AF. Only AF episodes lasting >15 minutes were considered.10 Atrial flutter and tachycardia were disregarded because of their different mechanisms.11 Amiodarone, either oral or intravenously administered, constituted the standard pharmacological treatment of AF. If medical therapy resulted in the reestablishment of sinus rhythm or in the control of ventricular rate and the patient was asymptomatic, the medication was continued for 6 weeks. In patients without successful rhythm cardioversion and with persistent AF, warfarin also was administered. At the end of this period and coincident with the first postoperative visit at the outpatient clinic, if the rhythm had reverted to sinus by a Holter ECG documentation, the medication (including warfarin) was discontinued. If the patient was still in AF, other cardioversion was considered. Symptomatic patients or failure of antiarrhythmic medications to control the ventricular rate led to cardioversion at any time during the postoperative period. All patients were discharged to a rehabilitation facility.
A survey questionnaire was given to patients. The following information was recorded: symptom change, occurrence of adverse cardiac and cerebrovascular events, cardiac-related hospital readmission after CABG operation, further cardiac interventions, medications, cardiac rhythm, and paroxysmal AF episodes. Causes of death were identified by a review of hospital records, death certificates, and autopsy reports. When these medical data were not available, causes of death were collected and obtained by telephone interviews of family members and/or confirmed or clarified by the general practitioners. All patients who entered the study had at least 1 follow-up visit at 3 months and/or 1 year. Follow-up was closed on April 30, 2007. Causes of death were classified as cardiac (acute myocardial infarction, heart failure, embolic) and noncardiac. Embolic events referred to hemorrhagic or ischemic strokes and pulmonary embolisms.
Data were described as mean and SD or median and interquartile range (25th to 75th percentile) for continuous variables and as number and percent for categorical variables. Median (25th to 75th percentile) follow-up was computed according to the inverse Kaplan-Meier method. The observation time extended from the date of discharge to the date of death or censoring. Preoperative and operative patient characteristics were compared according to the occurrence of postoperative AF by means of the Student t test or the Mann-Whitney U test and the Fisher exact test for continuous and categorical variables, respectively. The cumulative survival probability and 95% confidence intervals (CIs) were computed and plotted separately for patients with and without AF according to the Kaplan-Meier method. Mortality rates (95% CI) per 100 person-years were reported. The prognostic value of AF was evaluated by means of Cox regression. Hazard ratios (HRs) and 95% CIs were calculated. The proportional-hazard assumption was assessed with a test based on Schoenfeld residuals. All analyses were stratified on center. To account for the confounding effect of preoperative and operative characteristics, we further stratified the Cox model for AF by quintiles of the propensity score for AF. This was built by including in a logistic model variables with a value of P<0.10 from the univariable analysis as dependent variables and AF as the independent variable. Thus, the following variables were retained: center, age, dyslipidemia, chronic obstructive pulmonary disease, preoperative cerebrovascular accident, renal failure, left ventricular ejection fraction, postoperative intraaortic balloon pump, respiratory failure, stroke, and infection.
The prescription of warfarin at discharge also was included in the multivariable Cox model. Finally, in a secondary analysis, we evaluated the role of the AF occurrence during follow-up in a time-dependent Cox model including predischarge AF and warfarin while stratifying by center and quintiles of the propensity score. We then tested for its interaction with postoperative AF. To assess the stability of our results, we also performed a sensitivity analysis including patients lost to follow-up by assuming the same death incidence as that found in the patients who were followed up.
Cumulative incidence of death (95% CI) was computed for the following competing causes: cardiac versus other causes of death and embolic events versus heart failure versus acute myocardial infarction versus other causes of death. The role of postoperative AF in each cause of death was evaluated by fitting a Cox model stratified by cause of death and including the interaction of AF with each of the competing causes. Adjustment for confounders was performed as above.
Clinical data were tabulated with Microsoft Excel (Microsoft Corp, Redmond, Wash), and Stata 10 (Stata Corp, College Station, Tex) was used for computation. All tests were 2 sided, and a value of P<0.05 was considered statistically significant.
The overall incidence of AF during hospitalization in the whole study population was 31% (570 of 1832 patients). The profile for patients affected by postoperative AF was substantially different with regard to demographics, laboratory values, comorbid conditions, and perioperative factors compared with those without AF (Table 1⇓). Patients with AF were older with a lower left ventricular ejection fraction and a more severe profile of comorbidities. The use of β-blockers or other preoperative medications did not influence the prevalence of postoperative AF. Postoperatively, patients with AF had a higher prevalence of intraaortic balloon pump use (5% versus 3%; P=0.005), respiratory failure (12% versus 5%; P<0.001), strokes (3% versus 1%; P=0.006), and infections (14% versus 10%; P=0.016). They also had a longer length of stay in hospital (9.4±6.2 versus 8.1±5.2 days; P<0.001). As expected, patients with postoperative AF had a higher use of warfarin. Of 1832 patients, 26 died perioperatively; 19 (3.3%) deaths occurred in AF patients compared with 7 (0.5%) in patients without AF (P<0.001).
On discharge, 1806 patients were alive; 143 were lost to follow-up. Age, gender, and left ventricular ejection fraction distributions were similar to those of the remaining cohort (data not shown); 31 of them (21%) had had postoperative AF. All of these patients were unwilling to attend outpatient clinics, giving the distance from home as the reason for their nonattendance.
The remaining 1663 were followed up for a median of 51 months (25th to 75th percentile, 41 to 63); 126 of them died. The long-term mortality rates were significantly worse in patients with postoperative AF (2.99 per 100 person-years; 95% CI, 2.33 to 3.84; 61 deaths) compared with those without AF (1.34 per 100 person-years; 95% CI, 1.05 to 1.71; 65 deaths) with an HR of 2.26 (Table 2 and Figure 1). Similar estimates of increased risk of death for the AF patients were obtained after adjustment for confounding through the propensity score and after accounting for the prescription of warfarin at discharge (Table 2). In addition, patients on warfarin were shown to be at risk of poorer outcome (HR, 2.50; 95% CI, 1.33 to 4.41; P=0.005) according to the multivariable analysis.
In a secondary analysis, we assessed the independent predictive role of AF occurring during follow-up. A new episode of AF occurred in 119 patients (corresponding to an incidence of 1.3 per 100 person-years; 95% CI, 0.6 to 2.8). It did not appear to influence prognosis in the whole cohort (HR, 0.81; 95% CI, 0.30 to 2.20; P=0.68) in a model adjusted for predischarge AF and warfarin at discharge and stratified by center and quintiles of the propensity score. Given a significant interaction between predischarge and follow-up AF (P=0.007), separate models for the 2 groups of patients were fitted. AF that occurred during follow-up in the group without predischarge AF (but not in the group with predischarge AF) significantly increased the risk of death (HR, 4.44; 95% CI, 1.23 to 15.99; P=0.023).
In a sensitivity analysis including the patients lost to follow-up, with the incidence of death assumed to be the same as in the cohort that was followed up, results were unchanged (HR, 2.23; 95% CI, 1.57 to 3.16).
Eighty-two patients died of cardiac causes; 44 died of other causes. The cumulative incidence of death was computed as 7.2% (95% CI, 5.3 to 9.5) and 4.4% (95% CI, 3.0 to 6.2) for each of the 2 competing risks (Figure 2A). Patients with postoperative AF had a significantly higher risk of dying of cardiac causes, both in the univariable analysis and after adjustment for confounding, but not of dying of other causes (Table 2).
Among patients dying of cardiac causes, 54 had an embolic death, 11 died of heart failure, and 17 died of acute myocardial infarction. The cumulative incidence for these competing risks was computed to 4.2% (95% CI, 2.9 to 5.8), 1.2% (95% CI, 0.5 to 2.4), and 1.9% (95% CI, 0.9 to 3.5), as shown in Figure 2B). According to Cox regression, patients with AF were shown to be at higher risk of dying of embolism both in the univariable analysis and after adjustment for confounders (Table 2). The low number of deaths from acute myocardial infarction and heart failure did not allow us to reliably estimate the predictive role of AF in these cases.
Evidence that supports an association between AF after CABG surgery and late mortality is conflicting. Few data are available on the posthospital discharge course of patients with AF. It is generally believed that the risk of postoperative AF decreases substantially within the first month after surgery and is considered to be little more than a nuisance. However, most of the previous studies on the subject have focused on the perioperative period and have failed to include late survival data.1,3,5,6,12
Almassi and coworkers6 first demonstrated decreased survival rates in cardiac surgical patients with postoperative AF. At 6 months after surgery, mortality was significantly higher in AF patients compared with patients without AF (9.4% versus 4.2%).6 Villareal et al2 have recently reported an association between postoperative AF and the increased risk for early and late adverse outcomes in CABG patients. In this large case series, drawn from 6475 patients undergoing first isolated CABG, those affected by AF were at increased risk of death (odds ratio [OR], 1.5; 95% CI, 1.3 to 1.8). The adverse long-term mortality persisted after case-control matching (OR, 3.4; 95% CI, 1.6 to 7.5).2 In this study, cumulative survival rates at 1 and 4 years were 87% versus 94% and 74% versus 87% for the AF and non-AF populations, respectively.2 Mahoney et al7 reported similar data from >8500 isolated CABG patients. A significantly increased risk of death was observed in patients who developed postoperative AF compared with those who did not (HR, 1.2; 95% CI, 1.1 to 1.3).7 Similar results also were supported by our previous findings in a population of 7600 CABG patients who were followed up for 12 months after surgery.3 Patients affected by postoperative AF revealed an increased 1-year mortality (4.6% versus 2.0%), and AF was confirmed to independently predict late mortality (OR, 1.7; 95% CI, 1.2 to 2.5).3
However, all these studies showing a direct association between postoperative AF and mortality after cardiac surgery have been criticized for the presence of methodological flaws that may have compromised their validity. In several studies, incomplete matching led to the exclusion of large numbers of patients from the analysis.2,3,6,7 Despite attempts to account for confounding, it remains a distinct possibility that AF is associated with mortality because it occurs in patients with a severe profile of comorbidities.2,6,7,10 All these investigations did not provide plausible direct mechanisms linking postoperative AF with late mortality, and the causes of death were not investigated.2,3,6,7 Speculative causative mechanisms have been suggested, including hemodynamic ventricular compromise, occurrence of disabling strokes, and adverse proarrhythmic drug effects.2,3,6,8
Our study provides the first strong evidence that postoperative AF increases late mortality after cardiac surgery and is associated with attendant thromboembolic sequelae. The predictive power of AF persisted even after statistical adjustment for confounding variables through the propensity score. Postoperative AF increased the risk of dying of any cause by 2 to 3 times, tripled the risk of dying of cardiac causes, and quadrupled the risk of disabling embolic events. The Framingham study also examined the influence of AF on survival and reported similar data, with a significant reduction in survival in patients affected by the arrhythmia.13 Our survival curves resemble those of the Framingham study. Mortality rates at 1 and 10 years for patients with and without AF were 18% versus 5.9% and 61.5% versus 30%, respectively.13 Patients with AF had a 2.6- to 4.5-times increased risk of stroke, even after adjustment for other risk factors for stroke.13,14 A possible explanation was that AF created impaired hemodynamics from reduced ventricular filling and circulatory stasis in the left atrium and rendered patients susceptible to stroke and embolic events.14–16 In addition, reduced cardiac output and impaired cerebral circulation predisposed patients to cerebrovascular thrombosis causing noncardioembolic stroke.17
An interesting observation of our study was the effect of warfarin on postoperative outcome. The benefits of treatment with warfarin in reducing the risk of stroke in patients with AF have been well established by multiple randomized trials.18 Oral anticoagulation benefit accrued for either primary or secondary prevention, and the intensity of international normalized ratio of 2.0 to 3.0 has been recommended.18,19 In our study, however, warfarin therapy at discharge was independently associated with late mortality. In our centers, warfarin usually was given to patients suffering from left ventricular dysfunction or peripheral carotid disease who were intrinsically at a higher risk of dying. It might have been the case that some of these patients had an inadequate international normalized ratio range, which increased the risk of stroke. In the general population affected by AF, an inadequate international normalized ratio range in >40% of treated patients has been reported.20
The present study was limited by its observational nature with possible resultant confounding effects of measured and unmeasured variables on the role of postoperative AF. The advantages of this observational study were that a large population was made available for analysis and that the influences of real-world and current clinical practice were captured. Moreover, we attempted to adjust for a plethora of variables by means of the propensity score, although we acknowledge that unmeasured factors might have influenced the outcome. Moreover, studies that have used large observational databases have yielded information that was consistent with, and at times predicted, the results of randomized trials. Another limitation was the inability to obtain autopsy-certified data for all patients, a difficulty shared with other similar studies. However, causes of death were obtained whenever possible by a review of hospital records and death certificates collected from family members and confirmed by the patients’ general practitioners. A related potential weakness of this study was recall bias resulting from our reliance on self-reporting for follow-up cardiac rhythm data. Although commonly seen in clinical medicine, the use of questionnaires and clinical examination during follow-up may not have accurately captured paroxysmal episodes of AF and may potentially have underestimated the incidence of the arrhythmia. Another limitation of the study was the lack of complete data related to preoperative statin therapy. However, preoperative statins were not routinely prescribed in our centers at the time of study planning. These drugs were introduced only in the later years of the study. Consequently, we were not able to fully analyze the potential impact of preoperative statins on AF development.
Despite the limitations, our study supports the contention that postoperative AF is associated with an increased late mortality. The independent contribution of postoperative AF to mortality as a result of attendant embolic events also was demonstrated. Our results imply that appropriate antiarrhythmic and antithrombotic prophylaxis should be considered in patients after CABG.
We thank Martha Napoleone, Gianluca Belli, Andrea Pozzi, Vito Domenico Bruno, and Vincenzo Stringi for assistance with data collection.
Mariscalco G, Engstrom KG. Atrial fibrillation after cardiac surgery: risk factors and their temporal relationship in prophylactic drug strategy decision. Int J Cardiol. November 15, 2007. DOI: 10.1016/j.ijcard.2007.07.123. Available at: http://dx.doi.org/10.1016/j.ijcard.2007.07.123. Accessed June 24, 2008.
Hravnak M, Hoffman LA, Saul MI, Zullo TG, Whitman GR. Resource utilization related to atrial fibrillation after coronary artery bypass grafting. Am J Crit Care. 2002; 11: 228–238.
Benjamin EJ, Wolf PA, D'Agostino RB, Silbershatz H, Kannel WB, Levy D. Impact of atrial fibrillation on the risk of death: the Framingham Heart Study. Circulation. 1998; 98: 946–952.
Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation as an independent risk factor for stroke: the Framingham study. Stroke. 1991; 22: 983–988.
Yamanouchi H, Tomonaga M, Shimada H, Matsushita S, Kuramoto K, Toyokura Y. Nonvalvular atrial fibrillation as a cause of fatal massive cerebral infarction in the elderly. Stroke. 1989; 20: 1653–1656.
Atrial fibrillation (AF) is the most common arrhythmia in the general population and a frequent postoperative complication of cardiac surgery. AF is now widely recognized as a risk factor for stroke and other thromboembolic complications and heart failure. This leads to a substantial disease burden and significant medical costs. Although many studies have examined the impact of AF on the perioperative period, none has managed to present compelling data that support an independent association between postoperative AF and late mortality. Despite attempts to account for confounding, it remains a distinct possibility that AF is as associated with mortality as with potentially fatal comorbidities. Previously published studies have not provided plausible direct mechanisms that link postoperative AF with late mortality. In addition, the causes of death have not been investigated. In this study, we aimed to ascertain the impact of AF after coronary surgery on postoperative survival, assessing its prognostic role on cause-specific mortality. Our data have given perspective on the possible role of AF among these patients and the effect of embolic events in causing mortality.