Use of Intraoperative Transesophageal Echocardiography to Predict Atrial Fibrillation after Coronary Artery Bypass Grafting

Adult patients undergoing CABG surgery requiring cardiopulmonary bypass (CPB) from 1995 to 1997 were prospectively enrolled. Informed consent was obtained from all patients, in accordance with the hospital’s institutional review board. Patients were not eligible for the study if any of the following preoperative criteria existed: (1) cardiac surgical procedures in addition to the planned CABG surgery; (2) history of atrial or ventricular dysrhythmias or baseline rhythm other than sinus rhythm; (3) presence of antidysrhythmic medication other than digoxin or β blockers; and (4) conditions contraindicating the placement of a TEE probe. Patients were excluded from the analysis if there was intraoperative atrial fibrillation or an intraoperative change in the planned CABG procedure. No patient received pharmacologic prophylaxis for atrial fibrillation, with the exception of continuation of preoperative β-blocker therapy.

Patients were monitored prospectively from the time of surgery until hospital discharge. After placement of standard monitoring and induction of anesthesia, a pulmonary artery catheter was inserted, followed by placement of a 3.5–7 MHz phased array multiplane or biplane TEE probe (Acuson, Mountainview, CA).

Standard views of the heart were obtained using multiple imaging planes from 0–120° (multiplanar probes) or the transverse and the longitudinal imaging planes (biplanar probes). 15At the time of TEE examination, pulmonary capillary wedge pressure or pulmonary artery diastolic pressure was recorded. TEE examinations were conducted before sternotomy (pre-CPB) and after sternal closure (post-CPB) during periods of hemodynamic stability. Pulmonary vein and transmitral flow velocity patterns were obtained during brief periods of apnea and were measured during multiple cardiac cycles. The examinations were recorded on VHS videotape and evaluated later by two independent investigators who were blinded to the patients’ PostAF status. Final values represent the average of the two investigators’ results (each of which was the average of three individual measures). If the two investigators’ measurements differed by more than 5%, a third independent investigator performed the measurements, and the average of the three investigators’ values was used. The TEE views used to obtain the measurements included the midesophageal four-chamber view, the midesophageal two-chamber view, the midesophageal bicaval view, and the transgastric mid–short axis view. All Doppler recordings were conducted using electrocardiogram timing.

The measured parameters were left and right atrial size, left atrial appendage area (LAAA), left atrial (LA) appendage outflow velocity, pulmonary vein inflow velocities, mitral and tricuspid inflow velocities, 16and left ventricular ejection fraction (LVEF).

Electrocardiogram criteria for PostAF were defined as the presence of an irregular heart rate (R wave to R wave interval) with either the absence of electrical P-wave activity, or the presence of atrial fibrillatory waves. Patients were considered to have experienced an episode of PostAF if they had any electrocardiogram or telemetric documentation of PostAF for at least 1 min, as noted on the medical record, or if they required medical therapy for the dysrhythmia. Dysrhythmias were confirmed using 12-lead electrocardiogram when possible. Patients with dysrhythmias that could not be identified were not considered to have had atrial fibrillation. Patients were monitored until the time of discharge from the hospital using bedside electrocardiographic monitoring in the intensive care and step-down units, telemetry on the cardiac wards, and daily electrocardiogram when telemetry was discontinued.

Univariate analyses were conducted to identify potentially important predictors of risk of PostAF. This was performed with chi-square tests for binary predictors. Continuous variables were grouped into quartiles and analyzed using the chi-square test to determine evidence of trend. Apparent linear associations were confirmed in univariate logistic regression analyses. The variables with P < 0.2 in the univariate analysis were then entered in a multiple logistic regression analysis to study their joint influence on the occurrence of PostAF. The statistical significance of the factors in the multivariate model was tested using likelihood ratio tests comparing the model that included, with those that excluded, the relevant factor or factors. If a significant interaction term was identified, we report the significance level of the combined effect of the factors and their interaction. To further illustrate the joint effect of the factors and their interaction, we also report the probability estimates arising from our model for selected combinations of factor levels.

Of the 99 patients enrolled, 79 completed the study and had assessable examinations. Patients were excluded for the following reasons: intraoperative atrial fibrillation (n = 3) and missing preoperative data (n = 17). To be entered into the multivariate analysis, all of the measurements that were predefined had to be available in every patient.

The demographic data for the 79 subjects are presented in table 1. Of the 79 subjects studied, 28 (35.4%) experienced PostAF. The occurrence of PostAF in the 17 patients who were excluded for lack of preoperative data was 41%, which does not differ appreciably from the incidence in this study or in the general post-CABG population. Patients with PostAF had a median hospital length of stay of 10 days (range, 3–22 days) as compared with 8 days (range, 5–20 days) for those who did not have PostAF.

For dichotomous data, no demographic variable was significantly associated with PostAF. Of the continuous variables analyzed, only age was found to be a significant predictor for the development of PostAF (P = 0.001 by chi-square test for trend). Table 2shows the proportions of patients who developed PostAF for the demographic and preoperative variables studied.

None of the intraoperative parameters we studied was significantly associated with the development of PostAF. A comparison of mean values of specific intraoperative variables for patients who developed PostAF and those who did not is shown in table 3. A comparison of mean values of specific TEE variables for patients who developed PostAF and those who did not is shown in table 4.

The results of the univariate logistic regression analyses for parameters with P < 0.20 as predictors for PostAF are shown in table 5. These were age, pre-CPB LAAA, and post-CPB left upper pulmonary vein systolic to diastolic velocity (LUPVSD) ratio. Other TEE variables had P  values > 0.20.

Age, pre-CPB LAAA, and post-CPB LUPVSD ratio were entered in a multiple regression model to assess their combined influence on PostAF. In developing this model, it became apparent that the influence of increasing age was demonstrated primarily in association with increasing values of pre-CPB LAAA (table 5). This is evident from the statistically significant interaction term representing the product of pre-CPB LAAA and age (P < 0.05). In this model, the joint effect of pre-CPB LAAA and age is statistically significant (P < 0.005) as compared with a model that omitted the factors of age and pre-CPB LAAA and their synergistic effect. Post-CPB LUPVSD ratio is also significant, independently of these other factors (P < 0.05). To help interpret the meaning of this model, it was used to generate estimates of the probability of PostAF for various combinations of age, LAAA, and LUPVSD ratio, as shown in table 6. Note that the probability of PostAF as the LAAA increases from 3.0 to 4.0 cm is greater for the 75-yr-old patient than for the 65-yr-old patient at each level of LUPVSD ratio (table 6).

The incidence of atrial fibrillation after CABG surgery ranges from 10 to 40%. 1–3Multiple perioperative risk factors have been identified for the development of PostAF 1–14and the redevelopment of atrial fibrillation after conversion to sinus rhythm. 17–19Some of these risk factors include hypertension, CPB time, reduced LVEF, inadequate atrial protection, hypothermia, the type of surgical procedure performed, and advanced age. With the exception of advanced age, no consistent pattern of risk has emerged from these studies. TEE parameters have not been extensively evaluated.

By more specifically defining the population at increased risk for PostAF, we may be able to offer prophylactic pharmacologic therapy in a more discriminative manner. Class III antidysrhythmia agents, when given before CABG, have been shown to significantly reduce the occurrence of postAF. In randomized controlled trials, Daoud et al.  20administered amiodarone orally for 7 days before CABG, and Hohnloser et al.  21and Guarnieri et al.  22administered amiodarone postoperatively. Both showed significant reductions in the incidence of atrial fibrillation in the study group. Gomes et al.  23showed that prophylactic sotalol prevents the occurrence of postAF, and VanderLugt et al.  24showed similar results using prophylactic ibutilide. Because prevention seems possible, we sought to study preoperative, surgical, and TEE-derived data to identify parameters that would indicate an increased risk for developing PostAF.

The current study yielded an incidence of PostAF of 35.4%, which is consistent with other investigations. We found that age (P = 0.002), pre-CPB LAAA (P = 0.04), and post-CPB LUPVSD ratio (P = 0.03), when considered one at a time, were each associated with an increased risk of developing PostAF. The constellation of risk factors with the highest incidence of atrial fibrillation was high pre-CPB LAAA, in conjunction with older age, and low post-CPB LUPVSD ratio.

We did not find a significant relation between PostAF and a number of other factors that have been previously shown to be associated with PostAF. These include hypertension, CPB time, aortic cross clamp time, LA size, LA appendage outflow velocity, and LVEF.

In accordance with published literature, when age was considered alone, it was found to be a significant predictor of PostAF. 1,7,8,10,12The current multivariate analysis indicates that increased age is a risk for PostAF when in association with a large LAAA. This may be the result of age-related disease processes involving the myocardium (ischemia, valvular heart disease, congestive heart failure, and cardiomyopathy), which result in both mechanical and electrical atrial remodeling. 16,19,25These changes in atrial loading conditions manifest as atrial dilation and hypertrophy, 26which have been associated with PostAF. 19 

Several studies in outpatient populations with longitudinal follow-up have demonstrated an association between LA size and PostAF. 19,25,27,28However, the direction of causality is still speculative. 18,19,29Some investigators suggest that changes in LA loading conditions lead to LA enlargement and an increased predisposition to develop atrial fibrillation. 25,27Others propose that atrial fibrillation itself maybe the primary cause for the noted changes in LA size. 26,27 

The LA is a three-dimensional chamber, and it is difficult to quantify its size accurately with TEE because of the proximity of the posterior LA wall to the esophagus and near-field measurement inaccuracies. Measurement of the LAAA does not pose this specific limitation. Pre-CPB LAAA was associated with an increased incidence of PostAF in univariate analysis, the multivariate analysis indicating its strongest influence was evident in older patients. Because patients with a previous history of atrial fibrillation were excluded from our study, we can effectively rule out that atrial fibrillation led to enlarged LAAA in these patients. 29 

There are four potentially measurable velocity components in Doppler recordings of pulmonary venous flow. 30There are two ventricular systolic velocities, one ventricular diastolic velocity, and an atrial systolic velocity (flow reversal;fig. 1). With normal atrioventricular conduction, the two ventricular systolic velocities are often fused, and a single systolic velocity is seen in 70% of cases. 31We showed that after CPB, a decreased LUPVSD ratio was associated with an increased incidence of PostAF. In situations of left ventricular diastolic dysfunction, early closure of the mitral valve occurs, LA pressure increases, and a subsequent increase in the diastolic velocity with a reduction in systolic velocity occurs. This reduction in the systolic to diastolic forward flow ratio 32(decreased LUPVSD ratio) may be seen as an indicator of LA noncompliance and could be associated with an increased incidence of PostAF. Other conditions that result in a loss of LA compliance result in atrial remodelling 16,19and would have the same effect on pulmonary vein inflow velocities.

Previous studies have shown that a reduced LA appendage maximum outflow velocity is associated with atrial fibrillation and that this velocity increases when patients convert to sinus rhythm. 17Omran et al.  18showed that during sinus rhythm, peak emptying velocity of the LA appendage after cardioversion was significantly lower in those who would revert back to atrial fibrillation after cardioversion. However, our data and that of other investigators have not consistently shown LA appendage maximum outflow velocity to be a marker for atrial fibrillation. 33Because LA appendage maximum outflow velocity has been associated with recurrence of atrial fibrillation, it may be a result of reduced atrial function rather than a marker for new onset or post–cardiac surgery atrial fibrillation.

We were unable to demonstrate a significant association between LVEF and PostAF. Other studies have shown that a decreased ejection fraction is associated with an increased incidence of PostAF. 7,9Vaziri et al.  9showed that reduced left ventricular fractional shortening was a predictor of nonrheumatic atrial fibrillation in an elderly population. Mathew et al.  7identified congestive heart failure as an independent predictor of PostAF in patients undergoing coronary artery bypass surgery.

A decreased ejection fraction may be associated with ventricular and atrial noncompliance with resultant electromechanical dysfunction of the atrium leading to PostAF. 19This may have been masked in our study by the relative small number of patients with low LVEF and inclusion of only 6% of patients with congestive heart failure. A larger cohort may reveal an association between low LVEF and PostAF.

Several studies have found a relation between intraoperative techniques (duration of aortic cross clamp, CPB time, minimum temperature) and the incidence of PostAF. 5,6,34,35In the current study, we did not observe a significant relation between these factors and the incidence of PostAF. This may be a result of standard and consistent myocardial preservation techniques used among the surgeons at our institution.

Consistent with previous reports, PostAF is associated with longer overall hospital length of stay. It has been shown that pharmacologic therapy can reduce the occurrence of PostAF. 36However, before prophylaxis can be offered, those at risk need to be identified. We used our model to estimate the probabilities of PostAF for various combinations of patient parameters, as shown in table 6.

Several methodologic limitations of the current study exist and may make it difficult to generalize our findings to broader populations. The study was conducted on a specific population undergoing a relatively standard surgical protocol. In addition, the surgical techniques and myocardial preservation used in this study may differ from those at other institutions. A large number of potential risk factors were analyzed to identify important trends. We considered almost 50 potential factors and report the ones with significance levels less than P = 0.2. Thus, our findings should be considered the results of exploratory analyses and should be confirmed in a larger study.

Patients did not undergo any additional monitoring other than that which is used after CABG surgery. Thus, patients were not monitored with continuous electrocardiogram monitoring while in their regular hospital beds and may have had episodes of PostAF that went undetected. Nevertheless, the incidence of PostAF in our study is in keeping with that reported in the literature.

In conclusion, this study has identified large LAAA, increased age in conjunction with large LAAA, and low LUPVSD ratio as potentially important predictors of PostAF in CABG patients with no previous history of atrial dysrhythmia. If confirmed in larger studies, a model such as ours could be used to develop algorithms for identifying patients who are at high risk for developing PostAF. Prophylactic therapy could then be administered in the perioperative period with the aim of preventing PostAF, decreasing hospital length of stay, and reducing the overall cost of care.