Coronary artery disease (CAD), structural heart disease, and left ventricular (LV) dysfunction are among the factors that predispose to malignant ventricular arrhythmia (VA) (1). The incidence of VA and its impact on patients with CAD and LV dysfunction are well described ((2),(3),(4),5). Cardiac surgery (CS) exposes patients with a substrate for VA to various arrhythmic triggers, such as ischemia, reperfusion injury, hemodynamic changes, and electrolyte shifts, that could lead to post-operative VA (POVA). Of note, the reported incidence of POVA after CS is low, ranging from 0.95% to 3.2% ((6),(7),8). Thus, an analysis of POVA and its effect on patient outcomes requires a large cohort of patients undergoing CS. In this study, we sought to investigate the incidence, predictors, and outcomes of patients with POVA undergoing CS in a large cohort of patients.

This study was a retrospective cohort analysis to look for associations between POVA and long-term outcomes. We queried the Society of Thoracic Surgeons (STS) Adult Cardiac Database to identify consecutive patients undergoing CS at Emory University between January 2004 and July 2010. The entire cohort consisted of 14,720 patients. The extracted data included demographic variables, pre-existing comorbidities, operation type, and clinical outcomes. The Emory University Institutional Review Board approved the protocol in compliance with Health Insurance Portability and Accountability Act standards and the Declaration of Helsinki. The institutional review board waived obtaining individual informed consent before obtaining the data on these patients.

The primary study variable was the onset of POVA, defined as sustained ventricular tachycardia (VT) (lasting ≥30 s or requiring cardioversion) or ventricular fibrillation (VF) (242/248 patients met this criteria). The remaining 6 patients had frequent salvos of nonsustained VT that did not meet the 30-second duration, but the episodes were frequent, requiring intravenous antiarrhythmic drugs and transfer to an intensive care unit. POVA, as entered in the STS database, is defined as VT or VF with no specification as to which type is present. Thus, the charts of patients with POVA were reviewed to determine the mechanism of POVA (VT vs. VF) and the timing of POVA occurrence (≤48 vs. >48 h after surgery) when these variables were available from the medical record. We calculate the primary endpoint, the survival time after surgery, using the date of death of patients in the cohort, as identified by the Social Security Death Index (SSDI), a publicly available national database of death records extracted from the U.S. Social Security Administration's Death Master File Extract. The SSDI provided the date of death for each patient who died before the cutoff date of June 30, 2010, allowing computation of the Kaplan-Meier product-limit estimates and associated Cox regressions. Patients alive on this date were considered censored, and those undergoing operation after June 30, 2010, were excluded from survival analyses. Because the SSDI does not list the cause of death, the analysis describes all-cause long-term mortality.

To help determine the independent impact of POVA on survival, 40 risk factors of long-term mortality were identified and harvested from the STS database for use in a risk-adjusted analysis. Covariates included in the model are listed in (Table 1).

Missing data from the several selected variables included Caucasian race (n = 18, 0.1%), body mass index (n = 81, 0.6%), body surface area (n = 30, 0.2%), hemoglobin A1C (n = 1,765, 12.0%), height (n = 29, 0.2%), last creatinine level (n = 61, 0.4%), ejection fraction (EF) (n = 1,012, 6.9%), and number of diseased vessels (n = 1,200, 8.2%). Multiple imputation strategies were used in the final regression models to minimize the selection bias effect inherent in missing data settings.

Bivariate predictors of POVA were identified using 2-sample t tests and chi-square tests for numeric and categoric variables, respectively. These statistically significant variables were entered into an associative logistic regression, allowing the determination of independent predictors of POVA. Adjusted odds ratios (ORs) and 95% confidence intervals (CIs) were calculated using this analysis.

Unadjusted survival estimates were calculated using Kaplan-Meier product-limit estimation. Kaplan-Meier curves were generated to compare survival between patients with and without POVA. Two further stratifications were made in patients with POVA: (a) timing of POVA (> or ≤48 h) and (b) type of POVA (VF or VT, when available). To statistically evaluate the independent effect of POVA on long-term mortality, a multivariate Cox proportional hazards regression model was constructed that related survival time as a function of POVA while adjusting for 46 collected covariates to reduce confounding bias. Adjusted hazard ratios (AHRs) and 95% CIs were calculated for POVA and the 46 covariates.

All analyses were performed using SAS version 9.3 (SAS Institute Inc., Cary, North Carolina). No adjustments for multiple tests were performed.

In total, 14,720 patients were included in the analysis. The majority of patients were white men with a mean age of 61.7 years. More than 80% of the cohort had hypertension, one-third of patients had diabetes, and 36% of patients had the diagnosis of congestive heart failure. A total of 8,863 patients underwent isolated coronary artery bypass grafting (CABG) (60.2%), 3,238 patients underwent isolated valve surgery (22%), and 1,412 patients underwent combined CABG and valve surgery (9.6%). Half of all surgeries were performed off-pump (Table 2). POVA occurred in 248 patients (1.7%). (Table 2) shows the characteristics of patients with and without POVA. Patients with POVA were older (63.5 vs. 61.6 years), had lower mean EF (43.5 % vs. 51.3%), were more likely to have congestive heart failure (56.9% vs. 35.9%), and underwent emergent surgery (14.9% vs. 6.4%). The POVA group had more comorbidities with higher STS mortality risk score (9) (7.2% vs. 3.1%). Of note, patients with POVA were less likely to have undergone off-pump coronary artery bypass (27.0% vs. 50.6%) (Table 2). The mean length of stay for patients with POVA was double that of patients without POVA (16.3 vs. 8.03, p < 0.001).

After identifying the covariates associated with the occurrence of POVA in unadjusted bivariate analysis, a multivariable logistic regression model was performed to determine the independent predictors of POVA (Table 3). Six risk factors were identified as being independently associated with POVA. Older age was associated with 19.5% higher risk of POVA with each 10-year increase in age (OR: 1.019 per 1-year increase in age). The need for emergent surgery strongly predicted the occurrence of POVA (OR: 1.77, 95% CI: 1.10 to 2.86). Higher EF protected against POVA (OR: 0.974, 95% CI: 0.965 to 0.983). Each 10% increase in EF decreased the odds of POVA by 26%. Off-pump surgery was associated with a 59% reduction in the odds of POVA compared with patients undergoing cardiopulmonary bypass (OR: 0.41, 95% CI: 0.30 to 0.55) (Table 3). The presence of peripheral vascular disease (PVD) predicted the occurrence of POVA, whereas the presence of mild chronic obstructive pulmonary disease (COPD) seemed to be protective (Table 3).

POVA was associated with increased long-term mortality over a mean follow-up of 3.5 years. Patients with POVA had worse long-term survival than those without POVA, with a high risk of death in the POVA group during the first 6 postsurgical months (6-month survival of POVA = 59.8% vs. 93.8% for POVA-free group). This difference in survival persisted over time (Figure 11_gr1).

After adjusting for the covariates, POVA was independently associated with a 153% increase in long-term mortality compared with POVA-free patients (AHR: 2.53, 95% CI: 2.09 to 3.06). To better elucidate the effect of POVA on long-term mortality without the confounding effect of in-hospital mortality and early competing risks, a secondary analysis excluded patients who died before hospital discharge. This analysis showed that POVA independently predicted an increase in long-term mortality with a hazard ratio (HR) of 1.94 (CI: 1.46 to 2.60, p < 0.001). Other significant independent prognosticators of long-term survival (with hospital deaths excluded) are shown in (Table 4).

Charts of patients with POVA were reviewed to determine the mechanism of arrhythmia (VT vs. VF); these data were missing on 111 patients (45%) with POVA. Among patients with POVA, VF had a higher early hazard of death and produced worse long-term survival (unadjusted HR: 1.95, p = 0.007) (Figure 11_gr2). POVA occurring 48 h after surgery demonstrated worse early mortality risk but an equivalent survival at 6 years compared with POVA occurring within 48 h of surgery (unadjusted HR: 1.16, p = 0.31) (Figure 11_gr3).

The calculated mortality of patients with POVA within the first year after excluding in-hospital death reached 18.5% (27of 146 patients died within 1 year after hospital discharge) (Figure 11_gr4). We attempted to identify the characteristics of patients who are at high risk of death in the first year after surgery; patients who died at 1 year were more likely to have a lower EF (39.6% vs. 45.9% p = 0.08), increased age (67.9 vs. 62.6 years p = 0.044), and an index MI leading to surgery (37.0% vs. 23.5% p = 0.15). Because of the small number of patients, the EF and index MI did not reach statistical significance but did exhibit a strong statistical trend (p = 0.08 and 0.15 for EF and MI, respectively).

A total of 85 patients (34.3%) with POVA were treated with an antiarrhythmic drug (predominantly amiodarone). Twenty patients (8%) received an implantable cardioverter defibrillator (ICD) before discharge (a decision guided by an electrophysiologic study in some patients).

This study provides the largest analysis to date of the incidence, predictors, and outcomes of POVA in patients undergoing heart surgery. The 1.7% incidence of POVA in our population is consistent with that in prior reports ((6),(7),8). Steinberg et al. (6) followed 382 patients undergoing CABG and found that 12 (3.2%) developed POVA. Two other reports, one from Canada and one from England, enrolled 4,784 and 4,411 patients and found the incidence of POVA was 0.95% and 1.6% ((7),8), respectively. The potentially beneficial effect of myocardial revascularization might account for the observed low incidence of POVA in a population with structural heart disease ((10),11). Also, the favorable effect of valve surgery on LV loading conditions might reduce the proarrhythmic stress and stretch and associated proarrhythmic risk (12).

The current retrospective analysis identified 4 factors predisposing to POVA: increased age, lower EF, the need for emergent surgery, and PVD. Conversely, we found that off-pump surgery and mild COPD were associated with a lower chance of POVA. Advanced age repeatedly appears as a risk factor for long-term mortality after CS ((13),14). Two studies looking at risk of POVA after CS found that patients with POVA are older than patients without POVA ((6),8). Conversely, another study found that age <65 years at the time of surgery was a risk factor for POVA (7). LV dysfunction, expressed as a lower EF, also predicted POVA; the mean EF in patients with POVA in our study was 43.5% versus 51.3% for the group without POVA. A low EF also strongly predicted risk for POVA in prior studies that evaluated predictors of POVA after CS ((6),(7),8). The results of randomized clinical trials reinforce the role of EF as a major risk factor for predicting death in patients with or without associated coronary disease and myocardial infarction (MI) ((4),5). The need for emergent CABG also strongly predicted POVA in our patients. Emergency CABG generally carries a higher risk of in-hospital mortality and added requirements for hemodynamic support, such as the intra-aortic balloon pump and intravenous inotropes (15). Patients requiring emergency CABG typically have a higher CAD burden, intractable ischemia, LV dysfunction, and pre-operative hypotension or shock (15), all factors that contribute to the risk of VA.

Ascione et al. (7) reported a trend toward a reduction in POVA with the use of off-pump surgery (47% reduction in the relative risk, p = 0.1) in a center where 23% of CABG cases were performed off-pump. We observed an association between use of an off-pump surgical technique and a 58% relative reduction in POVA at a center aggressively using the off-pump approach, with more than 80% of isolated CABGs performed off-pump. This protective effect of off-pump CABG remained significant when using an intention-to-treat definition of off-pump CABG (i.e., surgeries that start as off-pump and had to be converted to on-pump are analyzed as off-pump). Avoiding cardiopulmonary bypass might protect against VA. A long on-pump time has been shown to predict POVA (8). The clinical consequences of cardiopulmonary bypass include systemic inflammation, oxidative stress, electrolyte shifts, release of vasoactive substances, and peripheral and cerebral emboli (16). Also, on-pump CABG is associated with a higher degree of myocardial injury evidenced by increased postsurgical troponin and creatinine kinase release ((17),18). Ischemia, myocardial injury, systemic inflammation, and electrolyte shifts may explain the higher risk of POVA observed with cardiopulmonary bypass.

Prior studies ((7),8) found an increased risk of POVA after heart surgery in women. In the current study, no gender-based difference in the incidence of POVA after heart surgery was observed. One possible explanation is that women in our series had higher rates of off-pump surgery than men (52.4% vs. 46%, p < 0.001), and its protective effect may have erased a gender-based difference.

The presence of PVD in this study is found to be predictive of POVA. PVD is associated with an increased risk of perioperative morbidity, including the risk of post-CABG atrial and VA ((19),20). It is also an independent predictor of mortality after CS (21). In addition, PVD is associated with multiple different comorbidities, such as LV dysfunction (22), which could predispose to VA. It is conceivable that PVD may signal severity of illness and the presence of multiple comorbidities that predispose to POVA.

Patients with moderate and severe COPD are at increased risk of post-CABG mortality ((23),24) and morbidity, including hospital length of stay, respiratory infections, and atrial fibrillation ((25),26). In our study, patients with mild COPD were less likely to have POVA. This may seem surprising given the association of COPD with worse outcomes after CABG. However, this association is mainly seen in patients with moderate or severe COPD, whereas mild COPD has not been demonstrated to predict worse outcome after CABG.

Prior studies have shown that POVA predicts higher (21.7% to 28.9%) in-hospital mortality compared with control (1.4% to 1.9%) ((7),8). Our data also show that the in-hospital mortality in patients with POVA was higher than in those without (31.5% vs. 2.9%, p < 0.001). Ascione et al. (7) noted that the detrimental effect of POVA on long-term mortality disappeared after correcting for the effect of in-hospital death. We did find that the adjusted HR decreased from an AHR of 2.53 (95% CI: 2.09 to 3.06) to an AHR of 1.94 (95% CI: 1.46 to 2.60) after excluding patients who died in-hospital; however, POVA still predicted an increased risk of long-term mortality. This single study analyzed the long-term outcome of 248 patients with POVA. The 3 prior reports analyzed data from a combined total of 126 patients with POVA. The larger number of patients followed in this study may provide adequate statistical power to detect the effect of POVA on long-term mortality.

We should stress that most deaths in patients with POVA occur in the hospital and within the first year after discharge (Figure 11_gr4). The observed 1-year mortality in the POVA group exceeded the yearly probability of death seen in patients enrolled in the MADIT-II (Multicenter Automatic Defibrillator Implantation trial-II) (4), a group with LV dysfunction associated with a previous MI. If one assumes that half of deaths resulted from VT or VF, as is generally the case in patients with CAD and LV dysfunction (27), these patients should be screened aggressively to look for markers of SCD with repeat ambulatory electrocardiography monitoring and/or provocative testing. Guidelines for an ICD as primary prevention of SCD currently exclude patients with an EF >0.35 and those within 90 days after undergoing surgical or percutaneous revascularization. Therefore, aggressive use of ambulatory electrocardiography monitoring to detect nonsustained VT or induction of VT during invasive electrophysiologic study would be needed to identify postsurgical patients who would be indicated to receive an ICD. Also, more aggressive use of a wearable defibrillator for a few months after discharge might protect those at risk during the first year after surgery.

We were able to compare the effect of timing of POVA on long-term outcome. POVA occurring within 48 h of CS resulted in similar long-term outcomes as POVA occurring >48 h after CS (p = 0.31). The prognostic significance of VA early after surgery has been unclear, given that the unstable electrical milieu early after cardiopulmonary bypass and reperfusion may produce transient electrical instability not associated with chronic risk. Likewise, patients with VA emerging early (<24 to 48 h) after MI have a better prognosis than patients with VA late after MI (>24 to 48 h) (28). Early-onset VA may result from acute ischemia or reperfusion injury, 2 potentially reversible conditions, and thus early POVA may have better outcomes than late POVA. Our data suggest that the timing of POVA (early vs. late) does not influence long-term outcome and challenge the traditional notion that early POVA has little if any long-term prognostic value and should be ignored after treating the acute episode. This analysis is limited by the availability of POVA timing data on only 154 of 248 patients (62%).

We also noted that patients with VF after CS have worse long-term survival than patients with VT (p = 0.007). This was mainly due to the strong impact of early death seen in patients with VF, an expected finding because VF results in immediate hemodynamic collapse whereas VT may support perfusion long enough to allow recognition and treatment before hemodynamic collapse. The availability of data on VA mechanism (55% of all POVAs) also potentially limits this analysis.

This is a single-center retrospective study analyzing the predictors and outcomes of patients with VA after CS, and thus it is subject to the same bias-based limitations affecting all such retrospective observational studies. The STS database was not designed to study the effect of POVA on long-term mortality or to determine the predictors of POVA. Therefore, the covariates included in the regression models might miss important factors that might have affected the association between POVA and long-term mortality. The SSDI reports all-cause mortality; therefore, data on cardiac and arrhythmic mortality are lacking. The charts of all patients with POVA were reviewed to confirm the accurate sustained and malignant nature of these arrhythmias. However, it is conceivable that some patients with wide complex tachycardia could have been misdiagnosed with VT, whereas the actual rhythm is supraventricular arrhythmia with aberrant conduction. In addition, some sustained but short episodes of hemodynamically stable POVA could have been missed. The bias reduction methods involve the construction of large models for POVA and long-term mortality that run the risk of poor estimation of effects via “over-fitting.” Furthermore, a small amount of missing data existed, and although sophisticated statistical methods were used to alleviate bias, the incomplete data collection could still result in less than optimal inferences. However, the large number of patients enrolled in this study and the long-term follow-up do suggest that POVA predicts long-term mortality after CS and that most of the mortality difference is secondary to a high rate of in-hospital death and within the first year after surgery in the POVA group.

In this large cohort of patients undergoing CS at Emory University, we have shown that older age, lower EF, and emergent surgery predisposes to POVA, whereas off-pump surgery is found to be protective against its occurrence. More important, POVA is associated with increased long-term mortality despite excluding in-hospital death.