Atrial Fibrillation After Coronary Artery Bypass Surgery
A Model for Preoperative Risk Stratification
Background—Atrial fibrillation (AF) occurs in 20% to 40% of patients after CABG. Identification of patients vulnerable for arrhythmia will allow targeting of those most likely to benefit from prophylactic therapy. The aim of the present study was to evaluate accuracy of a prospectively defined signal-averaged P-wave duration (SAPD) cutoff and additional preoperative characteristics for the prediction of AF after CABG.
Methods and Results—Patients undergoing elective isolated CABG were recruited to the present prospective study. SAPD was recorded in all patients. Filtered signals from 3 orthogonal leads were combined in a vector analysis, and total SAPD was measured preoperatively. Postoperative in-hospital AF occurred in 92 (28.2%) of 326 patients. Patients who developed AF were older (65.9 versus 61.7 years of age; P<0.0005) and had longer SAPD (158 versus 145 ms; P<0.0005) than non-AF patients. Incidence of AF increased in patients ≥75 years of age and increased progressively throughout the range of SAPD. Stepwise logistic regression analysis of preoperative variables identified that SAPD >155 ms (odds ratio, 5.37; 95% CI, 3.10 to 9.30; P<0.0005), advanced age (odds ratio, 1.53; 95% CI, 1.26 to 1.86 per 5-year increase in age; P<0.0005), and male sex (odds ratio, 2.88; 95% CI, 1.30 to 6.40; P<0.01) independently predicted AF. Prospectively defined SAPD >155 ms predicted AF with positive and negative predictive accuracy of 49% and 84%, respectively.
Conclusions—A combination of prolonged SAPD, advanced age, and male sex identifies patients at high risk for development of AF after CABG.
Atrial fibrillation (AF) occurs in 20% to 40% of patients after CABG.1 2 3 Incidence of arrhythmia has not changed, despite improvements in anesthetic and surgical techniques, and evidence suggests its incidence may be increasing.1
AF after CABG is self-limiting in most cases. However, even when AF is uncomplicated, its treatment requires additional medical and nursing time and a prolonged hospital stay.1 2 3 4 5 6 In a minority of cases, arrhythmia can cause hemodynamic compromise and increase risk of postoperative stroke.1 7 8 Consequently, AF after CABG leads to increased use of resources.2 3 4
Numerous studies have sought to identify predictors of AF after CABG. Their number is testament to the failure to prevent arrhythmia by prophylactic measures in unselected patients. Only increased age has consistently been associated with AF after CABG.1 2 3 9 10 11 Several workers, including those in our group, have demonstrated that prolonged preoperative signal-averaged P-wave duration (SAPD) is associated with AF after CABG.12 13 14 15
The present study was designed to evaluate prospectively accuracy of predefined SAPD for prediction of AF after CABG. The cutoff figure of 155 ms was derived from a previous study of 102 patients who underwent CABG.12 A secondary aim of the present study was to stratify patients at risk of AF on the basis of simple preoperative variables.
Of 358 eligible patients, 30 (8.4%) were excluded as a result of signal interference or high noise levels on SAPD recordings. Two patients who died within 72 hours of surgery were excluded. The remaining 326 patients form the basis of our results. Of these, 3 died before discharge: 1 each on postoperative days 8, 11, and 25. All 3 developed AF, on days 6, 1, and 2, respectively. Mean patient age was 63 years, and 268 (82%) were male. Preoperatively, 218 (67%) were taking β-blockers. After surgery, 8 (2.5%) returned to theater for sternal rewiring and 7 (2.1%) for bleeding.
Ninety-two (28.2%) patients developed AF, at 2.8±1.7 (mean; range, 0 to 11) days after operation. Five patients were noted to have brief (<30-s) episodes of irregular rhythm on Holter monitor; these episodes were not recorded on 12-lead ECG, and these patients were not included in the AF group. These patients did not develop AF during hospital stay, and their postoperative course was unremarkable. Demographic data are presented in Table 1⇓, and distribution of AF is illustrated in Figure 1⇓.
Low serum magnesium on the first postoperative day was not associated with AF. Levels <0.7 mmol/L were seen in 57% of patients who received cross-clamp fibrillation for myocardial protection, 22% who received cold-crystalloid cardioplegia, 14% who received whole blood, and 0% after CABG on the beating heart.
No differences were seen between AF and non-AF patients in left ventricular ejection fraction, Q waves on preoperative ECG, and left atrial diameter. Surgical techniques were similar between groups.
With univariate analysis, preoperative variables significantly associated with development of postoperative AF were age and SAPD. Use of preoperative β-blockers was more prevalent (70% versus 59%; P<0.05) in younger patients and those who remained in sinus rhythm. Weaker associations were noted between perioperative and postoperative variables of cardiopulmonary bypass time, units of blood transfused, and elevated postoperative plasma urea (>7 mmol/L) and creatinine (>125 μmol/L) concentrations. No difference was seen between groups in preoperative or postoperative magnesium concentrations. Postoperative β-blockers were used less commonly (26% versus 44%; P<0.005) in those who developed AF. β-Blocker withdrawal was significantly different (35% in the AF group versus 29% in those without AF; P=0.03). Postoperative hospital stay was significantly longer in patients with AF than in those without (9.2 versus 7.3 days; P<0.0005).
Advanced age was strongly associated with postoperative AF. Mean age was 65.9 years in the AF group compared with 61.7 years in the non-AF group (P<0.0005). Incidence of AF increased progressively in patients aged ≤75 years (Figure 2⇓), such that 1 of 29 (3.4%) patients aged 50 to 54 years versus 19 of 45 (42.2%) aged 70 to 74 years developed AF. Odds of developing AF increased 1.48 (95% CI, 1.24 to 1.77)-fold for each 5-year increase in age and 3.80 (2.00 to 7.23)-fold for those aged ≥60 years compared with patients <60 years of age.
No correlation existed between age and SAPD (R2=0.1; P=0.9). When incidence of SAPD >155 ms was documented within each 5-year age band, a small, nonsignificant difference was seen between groups. Those in the 55–to-69–years age band had lowest incidence of prolonged SAPD, 25.4%, and those in the 70-to-74-years age band, highest, 44.4% (P=NS).
Signal-Averaged P-Wave Duration
Mean SAPD was significantly longer (158 versus 145 ms; P<0.0005) in patients who developed AF after CABG. Incidence of AF increased progressively with increases in SAPD (Figure 3⇓), such that 3 of 50 (6%) patients with SAPD ≤130 ms developed AF compared with 8 of 13 (61.5%) with SAPD >175 ms. Odds ratio (OR) for AF was 2.11 (95% CI, 1.62 to 2.74) per 15-ms increase in SAPD and 4.95 (2.96 to 8.28) for SAPD >155 ms.
In the 64 patients in whom preoperative echocardiograms were recorded (18 developed postoperative AF), no association existed between left atrial size (measured at the level of the aortic valve leaflets in the parasternal long-axis view) and AF. Furthermore, no significant difference was seen between AF and non-AF groups in mean body weight (82.2 and 77.6 kg, respectively) or body mass index (27.7 and 26.5 kg/m2, respectively). We found no association between SAPD and either body weight or body mass index or with left atrial diameter. No significant difference was seen between the sexes in SAPD (148.9 ms for men versus 147.9 ms for women) and no significant difference in percentage of males (35.4%) versus females (39.7%) with SAPD >155 ms.
Stepwise logistic regression analysis of preoperative variables showed that age (P<0.0005), SAPD (P<0.0005), and male sex (P<0.01) were independently associated with development of AF (Table 2⇓). When all (preoperative and postoperative) variables were entered into analysis, >3 grafts (OR, 2.22; 95% CI, 1.17 to 4.21; P<0.02) and elevated postoperative plasma creatinine concentration (OR, 1.93; 95% CI, 1.04 to 3.60; P<0.04) were additionally associated with AF. β-Blockade use was not associated with AF preoperatively (P=0.098) or postoperatively (P=0.057).
Predictive Accuracy of Preoperative Variables
SAPD >155 ms predicted AF with a sensitivity of 63%, specificity of 74%, and positive and negative predictive accuracy of 49% and 84%, respectively. The combination of SAPD >155 ms and age ≥60 years increased positive predictive accuracy to 57%, with a negative predictive accuracy of 82%. Age of ≥60 years alone had a positive and negative predictive accuracy of 35% and 87%, respectively. When combined with male sex, respective values were 39% and 87%.
Table 3⇓ stratifies patients by SAPD >155 ms and aged ≥60 years. Table 3⇓ shows that the incidence of AF after CABG for patients aged <60 years with SAPD ≤155 ms was 6.8%, versus 56.8% for patients aged ≥60 years with SAPD >155 ms.
Similarly, Table 4⇓ is a risk stratification table that includes gender. Interestingly, 0 of 35 women with SAPD ≤155 ms developed AF, irrespective of age, whereas 11 of 23 (47.8%) women with SAPD >155 ms developed AF. Highest incidence was in males ≥60 years with SAPD >155 ms, who had a 58.8% chance of AF after CABG.
The present study confirms both high incidence of AF after CABG and its strong association with increases in age and SAPD. The present study is the largest to evaluate SAPD for the prediction of AF after CABG and the first to validate a cutoff value prospectively defined from a study of patients undergoing CABG. We have described incidence of AF in a population undergoing elective, isolated CABG and quantified the risk from SAPD, age, and sex, thereby to allow preoperative identification of vulnerable patients.
Despite improvements in surgical and anesthetic techniques, incidence of AF after CABG remains stubbornly unchanged. Little doubt exists that this is, at least in part, a reflection of increased age of patients undergoing surgical revascularization,1 16 itself a consequence of the aforementioned improvements.
Patients who develop AF have a prolonged postoperative hospital stay compared with patients who remain in sinus rhythm; this finding was confirmed in the present study. Recent evidence shows that prolonged hospitalization is attributable to arrhythmia rather than clinical characteristics of AF patients.6 The financial burden that results from prolonged hospitalization is considerable.2 3 4
We failed to confirm our previously reported association of AF with postoperative hypomagnesemia. One reason could be the different strategies of myocardial protection used in the 2 studies. In the earlier study12 post hoc analysis revealed 59% of patients received cross-clamp fibrillation and 41% cold-crystalloid cardioplegia. Patients in the former group had a significantly lower serum magnesium level on the first postoperative day (P<0.005; unpublished data). In the present study, although cross-clamp fibrillation was again associated with postoperative hypomagnesemia, most patients received whole-blood cardioplegia supplemented with magnesium. However, prevalence of AF in the 2 studies was similar, which suggests that postoperative magnesium is not causally related to arrhythmia development.
Advanced age increases risk of AF in the general population.17 18 Therefore, it is not surprising that incidence of AF after CABG also increases with age. Dilatation and fibrosis of the atria have been shown to increase with age,19 with a loss of side-to-side electrical coupling between groups of atrial muscle fibers.20 Consequent slowing of electrical conduction within the atria provides a substrate for arrhythmogenesis. Of interest, the increase in AF was not sustained in patients >75 years of age, a finding previously reported in patients >80 years of age.3 Mathew et al postulated that the increase in atrial connective tissue with age and resulting nonuniform anisotropic conduction could be maximal by the eighth decade of life in humans. Although numbers of patients within this age group were relatively small, the finding is important, given that the mean age of patients undergoing CABG continues to increase.1 16
AF is generally accepted to be reentrant in origin.21 22 23 Therefore, sustained AF requires that the depolarizing wave fronts continuously encounter excitable tissue, a circumstance favored by slow atrial conduction and a short atrial refractory period.24 SAPD has been established to be prolonged in patients who develop AF after CABG, and this is thought to reflect slow atrial conduction in patients with the arrhythmia substrate.25 26 27 28 29 Absence of a relation between SAPD and age or body weight in the present study suggests that the former is an independent marker of the arrhythmia substrate.
Prolonged SAPD merely identifies patients with the arrhythmia substrate. The trigger for development of AF is probably multifactorial. Cox proposed that vulnerability to development of AF after cardiac surgery is increased by nonuniformity of local refractory period distribution within the atrium.30 This dispersion of atrial refractoriness may be promoted by perioperative factors such as atrial ischemia, imbalance in autonomic tone, or electrolyte imbalance. That an additional trigger is necessary to initiate AF in those with the arrhythmia substrate is a possible explanation for the moderate positive predictive accuracy of SAPD. Those without the arrhythmia substrate and, hence, at low risk for AF after CABG are identified more accurately.
Several studies have found an increased incidence among males,1 2 3 9 whereas others have reported no difference.6 31 32 In our study, the number of women was small, and any conclusion should be treated with caution. We found no differences on univariate analysis, but male sex was an independent predictor of AF after CABG on multivariate analysis and improved the accuracy of SAPD in identification of vulnerable patients.
Left Atrial Size
We did not find a relation between AF and left atrial size or between SAPD and left atrial size in a subset of patients. This is consistent with results from other studies.3 33 Given the strong association between left atrial size and AF in the general population, this result is surprising and may indicate that a different mechanism underlies arrhythmia induction in the postoperative group.
Use of β-blockers preoperatively protected against arrhythmia on univariate analysis, but this effect was lost when stepwise logistic regression with other significant preoperative variables was performed. Previous reports suggest that β-blockers suppress AF after surgery.34 35 However, in our study, β-blocker use was confounded by age in multivariate analysis. Younger patients were given β-blockers more commonly than older patients, and age was more strongly associated with AF. We emphasize that the present study was not designed to establish drug efficacy. Our sole aim was to identify preoperative variables in risk stratification.
Pharmacological intervention to reduce AF incidence in unselected patients has proved disappointing. Daoud et al36 reported amiodarone to be of benefit, but the study had major limitations and the in-hospital benefit was marginal.37 The ability to identify patients at increased risk for AF after CABG will increase the statistical power of studies to detect a meaningful reduction in AF in the treated group. Conversely, recognition of low-risk patients indicates those who will not benefit from preventive therapy and allows better planning of hospital resources in patient management after cardiac surgery. Given that the additional cost of AF after CABG in the United States is estimated to be $153 million,3 information on likely risk and possible increased hospital stay is useful to both providers and purchasers of health care.
Signal-averaged ECG is a noninvasive test that takes 20 minutes to record. The machine is portable and allows bedside recording, and it is thus suitable for outpatient use. By use of additional preoperative clinical variables, we have significantly improved the ability of high-resolution ECG to identify patients at risk of AF after CABG.
As with all similar studies, one limitation of the present study is the method used to record arrhythmias. We continuously monitored rhythm on all patients for ≥72 hours and recorded 12-lead ECGs if arrhythmia was suspected clinically. Our AF prevalence of 28% is within the 25% to 30% range reported in the majority of such studies.
The present study is a prospective study of patients undergoing elective CABG at 2 regional cardiac centers in the United Kingdom. Patients participating in other trials were excluded, as were those undergoing concomitant valve surgery, in AF or with a history of AF, or receiving class I or III antiarrhythmic agents. All patients gave informed consent, and the study was approved by the local ethics committee for each of the participating centers.
Preoperative SAPD was measured in all patients. In addition, the following data were recorded prospectively. Preoperative data included age, sex, medications, presence of Q waves on 12-lead ECG, coronary anatomy, left ventricular ejection fraction on angiogram, hemoglobin, urea, electrolytes, and magnesium concentration. Echocardiograms were performed when possible (n=64) to assess left atrial size. Perioperative data included number of distal anastomoses, use of internal mammary artery graft, aortic cross-clamp time, cardiopulmonary bypass time, method of myocardial protection, and operator status. Postoperative data included units of blood transfused, hemoglobin, urea, electrolytes, and magnesium concentration on first postoperative day; duration of postoperative hospital stay; and drugs at discharge.
The method of recording and analyzing P-wave signal-averaged ECG has previously been described.12 38 ECGs were recorded and analyzed by a single individual at each center using a Predictor IIc signal averaging system (Arrhythmia Research Technology Inc). After preparation of the skin, silver-silver chloride electrodes were applied in an orthogonal arrangement. QRS was used as the trigger for the signal-averaging process. To expose the P wave and PR segment, the fiducial point was shifted to the extreme right of the 300-ms window. A P-wave template was manually selected, and P-wave complexes that did not match the template with a 99% correlation coefficient were rejected. This ensured that any variation in the P-wave complex due to surface noise, ectopic activity, or autonomic activity was not averaged. Recordings were terminated when 500 beats had been accepted for averaging. If noise exceeded 0.5 mV, the recording was repeated until the criterion was met. A least-squares-fit filter with a window width of 100 ms (high-pass cutoff, 29 Hz) was applied for data analysis. P-wave complexes of the X, Y, and Z bipolar leads were combined into a vector magnitude with the root-mean-square of each lead. P-wave onset and offset were determined manually in the spatial magnitude, and P-wave duration was calculated preoperatively. P-wave onset was defined as the first atrial deflection from baseline noise level and offset as return of the atrial signal to baseline or onset of the QRS complex, whichever was earlier.38
Reproducibility was assessed by blinded remeasurement of 20 randomly selected P waves by the original investigator (for intraobserver reproducibility) and by a second investigator (for interobserver reproducibility). Mean intraobserver difference in the absolute SAPD measurement was +1.2±6.7 ms, which represented a coefficient of reproducibility of 13.4, or 8.7% when expressed as percentage of mean value for 2 measurements.39 Interobserver values were +3.0±10.8 ms, or 21.5% and 14%. Intraobserver agreement for greater than or less than the 155-ms cutoff occurred in 18 of 20 cases (90%; κ=0.8; P<0.0005), and agreement between observers occurred in 17 of 20 cases (85%; κ=0.7; P=0.002).
All patients underwent isolated elective CABG. Material for vein grafts was harvested from saphenous vein. Internal mammary artery grafts were used when appropriate. Four patients underwent CABG surgery performed on a beating heart. Myocardial protection was achieved with blood cardioplegia supplemented with magnesium (n=243), cold-crystalloid cardioplegia (n=9), or intermittent cross-clamp fibrillation (n=72) according to operator preference. Core temperature was allowed to decrease to 28°C to 32°C. Each patient was rewarmed to 37°C as cardiopulmonary bypass was discontinued.
Postoperatively, patients were managed in the intensive care unit before transfer to the high-dependency unit and main ward. ECG was monitored continuously by bedside monitor. Patients transferred to the main ward before the third day underwent telemetry or ambulatory ECG to allow continuous monitoring for 72 hours postoperatively. Thereafter, patients were examined daily, and an ECG was performed whenever AF was suspected clinically. Nursing staff members were asked to record tachycardias noted during monitoring by 12-lead ECG. Study end point was any episode of documented (on 12-lead ECG) postoperative in-hospital AF, defined as an irregular rhythm with no identifiable P waves. Prophylaxis against postoperative AF was not used routinely, and its occurrence was managed by the attending physician. Prescription of drugs was not influenced by study investigators.
Data were stored electronically and analyzed by use of SPSS for Windows statistical software. Univariate analysis by Student’s unpaired t test was performed for continuous data with a normal distribution, by Mann-Whitney U test for nonnormal data, and by χ2 test for categorical variables. Stepwise logistic regression was performed to investigate the effect of possible confounders on SAPD. Variables with univariate significance ≤0.1 were entered into the regression. Preoperative variables were considered before a separate analysis of all variables. For purposes of stepwise logistic regression, nonnormal data were dichotomized. Predictive accuracy of the prospectively defined SAPD >155 ms was determined. ORs for AF were calculated for these preoperative variables: SAPD, age, and male sex. Risk tables were generated from incidence of AF in different patient groups.
The present study was supported by a British Heart Foundation project grant (No. 96/188). Dr Zaman was a Fulbright Scholar supported by an International Fellowship from the British Heart Foundation.
The Methods section of this article can be found at http://www.circulationaha.org
- Received July 8, 1999.
- Revision received October 12, 1999.
- Accepted October 19, 1999.
- Copyright © 2000 by American Heart Association
Aranki SF, Shaw DP, Adams DH, Rizzo RJ, Couper GS, VanderVliet M, Collins JJ Jr, Cohn LH, Burstin HR. Predictors of atrial fibrillation after coronary artery surgery: current trends and impact on hospital resources. Circulation. 1996;94:390–397.
Mathew JP, Parks R, Savino JS, Friedman AS, Koch C, Mangano DT, Browner WS, for the Multicenter Study of Perioperative Ischaemia Research Group. Atrial fibrillation following coronary artery bypass graft surgery: predictors, outcomes, and resource utilization. JAMA. 1996;276:300–306.
Zaman AG, Alamgir F, Richens T, Williams R, Rothman MT, Mills PG. The role of signal averaged P wave duration and serum magnesium as a combined predictor of atrial fibrillation after elective coronary artery bypass surgery. Heart. 1997;77:527–531.
Steinberg JS, Zelenkofske S, Wong SC, Gelernt M, Sciacca R, Menchavez E. Value of the P-wave signal-averaged ECG for predicting atrial fibrillation after cardiac surgery. Circulation. 1993;88:2618–2622.
Stafford PJ, Kolvekar S, Cooper J, Fothergill J, Schlindwein F, deBono DP, Spyt TJ, Garratt CJ. Signal-averaged P-wave compared with standard electrocardiography or echocardiography for prediction of atrial fibrillation after coronary bypass grafting. Heart. 1997;77:417–422.
Noyez L, Janssen DP, van Druten JA, Skotnicki SH, Lacquet LK. Coronary bypass surgery: what is changing? Analysis of 3834 patients undergoing primary isolated myocardial revascularisation. Eur J Cardiothorac Surg. 1998;13:365–369.
Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation as an independent risk factor for stroke: the Framingham Study. Stroke. 1991;22:983–988.
Davies MJ, Pomerance A. Pathology of atrial fibrillation in man. Br Heart J. 1972;34:520–525.
Spach MS, Dolber PC. Relating extracellular potentials and their derivatives to anisotropic propagation at a microscopic level in human cardiac muscle: evidence for electrical uncoupling of side-to-side fiber connections with increasing age. Circ Res. 1986;58:356–371.
Moe GK. On the multiple wavelet hypothesis of atrial fibrillation. Arch Intern Pharmacodyn Ther. 1962;140:183–188.
Simpson Jr RJ, Amara I, Foster JR, Woelfel A, Gettes LS. Thresholds, refractory periods, and conduction times of the normal and diseased human atrium. Am Heart J. 1986;116:1080–1090.
Kumagai K, Akimitsu S, Kawahira K, Kawanami F, Yamanouchi Y, Hiroki T, Arakawa K. Electrophysiological properties in chronic lone atrial fibrillation. Circulation. 1991;84:1662–1668.
Suttorp MJ, Kingma JH, Tjon Joe Gin RM, van Hemel NM, Koomen EM, Defauw JA, Adan AJ, Ernst SM. Efficacy and safety of low and high dose sotalol versus propranolol in the prevention of supraventricular tachyarrhythmias early after coronary artery bypass operations. J Thorac Cardiovasc Surg. 1990;100:921–926.
White HD, Antman EM, Glynn MA, Collins JJ, Cohn LH, Shemin RJ, Friedman PL. Efficacy and safety of timolol for prevention of supraventricular tachyarrhythmias after coronary artery bypass surgery. Circulation. 1984;70:479–484.
Ali IM, Sanalla AA, Clark V. Beta-blocker effects on postoperative atrial fibrillation. Eur J Cardiothorac Surg. 1997;11:1154–1157.