Risk of atrial fibrillation (AF) after coronary artery bypass grafting (CABG) is high, yet the effectiveness of guideline-recommended preoperative prophylaxis in clinical practice remains uncertain. We determined the utilization and variation of preoperative AF prevention and assessed the comparative effectiveness of alternative drugs using the Society of Thoracic Surgeons multicenter Contemporary Analysis of Perioperative Cardiovascular Surgical Care (CAPS-Care) registry. Among 2,177 patients who underwent high-risk CABG and/or valve surgery, the mean age was 71 ± 9, 66% were men, 26% had chronic lung disease, and 21% had cerebrovascular disease. Overall use of AF prophylaxis was 84% and varied across sites (range 52% to 100%). The most common preventive agents were beta blockers (72%), followed by calcium antagonists (17%). Postoperatively, 30% (n = 646) developed AF at a median of 2 (25th to 75th percentiles: 1 to 3) days after surgery. Increasing age, height, white race, body mass index >35, New York Heart Association class IV heart failure, preoperative dialysis, and concomitant aortic valve replacement were associated with greater odds of postoperative AF (p <0.05 for all). Preoperative amiodarone use was associated with a trend to reduction of postoperative AF (26%, adjusted odds ratio 0.72 [95% confidence interval 0.51 to 1.00], p = 0.052). After adjustment, the odds of postoperative AF were not statistically different across agents. In conclusion, use of AF prophylaxis before surgery varied significantly. In this high-risk population, we were unable to demonstrate that any of the commonly used preventive agents were associated with lower rates of AF compared with alternatives or no treatment.
Postoperative atrial fibrillation (AF) is common, affecting 25% to 40% of patients after coronary artery bypass grafting (CABG). Not only is this dysrhythmia highly prevalent in CABG patients, but it is also associated with considerable morbidity and mortality, including an increased risk of stroke, renal failure, and heart failure. Postoperative AF is also costly, increasing length of stay and elevating total health care resource requirements. Numerous trials have examined interventions for the prevention of postoperative AF; however, their applicability has been largely limited to comparison of single-treatment strategies. Few head-to-head active comparator trials have been done, and these were often conducted in selected patient populations and/or centers. The goal of this study was to examine variation in the use of preoperative pharmacotherapy and its comparative real-world effectiveness for the prevention of postoperative AF in a large multicenter population undergoing high-risk CABG.
The Society of Thoracic Surgeons National Cardiac Database (STS NCD) is a nationwide quality improvement registry of cardiac surgery. The details of the STS NCD participation and data collection have been previously described. The Contemporary Analysis of Perioperative Cardiovascular Surgical Care (CAPS-Care) is a subregistry within the STS Adult Cardiac Surgical Database that collects additional information in patients who undergo high-risk or urgent CABG with or without concomitant valvular or AF surgery. CAPS-Care was designed to study the characteristics, management, and outcomes of patients following high-risk CABG in the United States. Patients were eligible if they had a preoperative ejection fraction <40% or age ≥65 years, with either diabetes mellitus or an estimated glomerular filtration rate <60 ml/min per 1.73 m 2 . Patients were excluded if they were ≤18 years or if they had an emergent/salvage operation or preoperative cardiogenic shock. For the purpose of this analysis, patients in AF before surgery and those patients who underwent concomitant AF correction surgery were excluded.
Data for the CAPS-Care study were collected from May 11, 2006, through December 31, 2006 at 48 STS NCD registry sites. A total of 2,177 patients were eligible and entered into the analysis data set. Data elements included demographics, medical history, clinical presentation, medical therapy before hospitalization and in the hospital, hospital course including procedures, in-hospital outcomes, and discharge disposition. Data elements from the patients’ corresponding STS NCD files were also used in the analysis, consistent with previous CAPS-Care analyses.
Postoperative AF was defined and recorded on the case report form as any sustained new onset AF or atrial flutter requiring an intervention that occurred during the postoperative period before discharge. Patients who were in AF at baseline were excluded. The incidence of postoperative AF was assessed as a function of preoperative drug therapy. Prespecified treatment categories included overall, beta blocker, calcium antagonist, sotalol, amiodarone, other antiarrhythmic, none. Medication dosage was not available. Prespecified operative characteristics included an urgent indication for surgery, CABG only, CABG and aortic valve replacement, CABG and mitral valve repair, CABG and mitral valve replacement, off-pump surgery, perfusion time, and cross-clamp time.
Baseline characteristics and operative characteristics were summarized according to the occurrence of postoperative AF using medians and 25th to 75th percentile interquartile range for continuous variables and frequency and percentage for categorical variables. Preoperative AF prophylaxis was defined according to the prespecified treatment categories.
Variation in preoperative prophylaxis was described at the site level (n = 48). To display the results graphically, use of preoperative AF prophylaxis was plotted against the number of patients at each site. We superimposed lines representing 95% binomial prediction limits, indicating the range of results that would normally occur as a result of random statistical variation for a hospital whose true frequency of using an agent was equal to the mean for the whole population.
We explored the association between baseline clinical characteristics and postoperative AF. Logistic regression modeling was used to estimate the risk of postoperative AF as a function of baseline patient variables. Using backward selection with a significance criterion of p = 0.05, independent predictors were identified from a list of covariates including previously reported predictors of postoperative AF and potentially relevant variables ( Appendix ). The discrimination of the full and the reduced models were assessed by the use of Harrell’s C-index. The enhanced bootstrap was used to estimate the bias (i.e., overestimated C-index) due to model overfitting in the original sample. Risk adjusted odds ratios (ORs) of covariates in the reduced models were then estimated. Robust sandwich variance estimates were used to obtain 95% confidence intervals (CIs) to account for statistical dependence of patients within sites.
The association between preoperative prophylaxis and postoperative AF was assessed using the unadjusted logistic model containing treatments that indicated the usage of amiodarone, sotalol, other antiarrhythmic therapy, beta blocker, or calcium antagonist therapy, and the risk-adjusted model with both treatments and a set of relevant covariates ( Appendix ). Unadjusted and adjusted ORs for using a specific agent versus not using that agent were reported. Combinations of agents (any antiarrhythmic + beta blocker, any antiarrhythmic + calcium antagonist) were examined in a separate adjusted model with the same covariates.
Among 2,177 patients in the CAPS-Care registry, 30% (n = 646/2,177) had sustained AF after surgery. The median time to AF onset was 2.0 days (25th to 75th percentiles: 1.0 to 3.0). Baseline characteristics according to the occurrence of postoperative AF are shown in Table 1 . Those who developed postoperative AF were older, more commonly hypertensive, and had worse renal function. There were no significant differences in the sex or body mass indices between those patients with and without postoperative AF. More patients with New York Heart Association (NYHA) class IV heart failure (vs NYHA class I to III) developed postoperative AF, although the median left ventricular ejection fraction was higher in those with postoperative AF (49% vs 43%).
|Variable||Postoperative AF||p Value|
|No (n = 1,531)||Yes (n = 646)|
|Age (yrs)||71 (65–76)||74 (69–79)||<0.0001|
|Female||517 (34%)||217 (34%)||0.9362|
|Caucasian||1,344 (88%)||588 (91%)|
|Black||73 (5%)||16 (2%)|
|Hispanic||34 (2%)||12 (2%)|
|Asian||15 (1%)||8 (1%)|
|Native American||10 (1%)||1 (1%)|
|Other||47 (3%)||20 (3%)|
|Hypertension ∗||1,265 (83%)||556 (86%)||0.0507|
|Diabetes||819 (53%)||343 (53%)||0.8648|
|Hypercholesterolemia †||1,209 (79%)||469 (73%)||0.0025|
|Current tobacco use||322 (21%)||78 (12%)||<0.0001|
|Cerebrovascular disease||308 (20%)||159 (25%)||0.0234|
|Previous stroke||221 (14%)||115 (18%)||0.0498|
|Peripheral vascular disease||292 (19%)||139 (22%)||0.2109|
|COPD (FEV 1 <60%)||206 (13%)||78 (12%)||0.3823|
|Preoperative dialysis||26 (2%)||23 (4%)||0.0079|
|Midwest||630 (41%)||254 (39%)|
|Northeast||163 (11%)||70 (11%)|
|South||471 (31%)||191 (30%)|
|West||267 (17%)||131 (20%)|
|Previous cardiac interventions|
|Any previous cardiovascular surgery||139 (9%)||49 (8%)||0.2571|
|Previous coronary bypass||109 (7%)||39 (6%)||0.3507|
|Any previous PCI||289 (19%)||106 (16%)||0.1527|
|Previous CABG or PCI||353 (23%)||136 (21%)||0.3061|
|Previous cardiac valve surgery||16 (1%)||7 (1%)||0.9425|
|Preoperative medical therapy|
|Beta blocker||1,061 (69%)||446 (69%)||0.9041|
|ACE inhibitor||688 (44%)||286 (44%)||0.7755|
|Aspirin||1,123 (72%)||471 (72%)||0.6786|
|Warfarin||15 (1%)||10 (2%)||0.3168|
|Lipid-lowering therapy (any)||581 (38%)||252 (39%)||0.5603|
|Statin||562 (37%)||242 (37%)||0.6139|
|Thienopyridine||89 (6%)||36 (6%)||0.5978|
|Preoperative clinical data|
|Body mass index (kg/m 2 )||28 (25, 32)||28 (25, 32)||0.5809|
|Heart rate (beats/min)||70 (60, 79)||68 (60, 77)||0.0446|
|Preoperative heart rate ≥100||25 (2%)||15 (2%)||0.3546|
|Systolic blood pressure (mm Hg)||130 (112–145)||130 (112–145)||0.7611|
|Diastolic blood pressure (mm Hg)||62 (56–72)||60 (55–70)||0.1862|
|Hemoglobin (mg/dl)||13 (11.6–14.2)||12.7 (11.4–14.1)||0.0251|
|Estimated GFR (ml/min/1.73 m 2 )||59 (49, 79)||57 (45–75)||<0.0001|
|LVEF||43 (30, 56)||49 (35–60)||0.0057|
|Heart failure||422 (28%)||203 (31%)||0.0712|
|I||167 (11%)||60 (9%)|
|II||318 (21%)||139 (22%)|
|III||682 (45%)||270 (42%)|
|IV||357 (23%)||174 (27%)|
|Sinus||1,397 (91%)||595 (92%)|
|Paced||56 (4%)||16 (2%)|
|Other||78 (5%)||36 (6%)|
|Preoperative CHADS 2 score||0.0002|
|0||77 (5%)||16 (2%)|
|1||322 (21%)||106 (16%)|
|≥2||1,132 (74%)||524 (81%)|
∗ Hypertension was defined in the STS registry as diagnosed and treated with diet and/or exercise, systolic blood pressure >140 or >90 mm Hg diastolic on ≥2 occasions, or current treatment with an antihypertensive medication.
Table 2 shows the operative data according to the occurrence of postoperative AF. Among this cohort, 1,706 patients (78.4%) underwent isolated CABG, and 1,059 patients (48.6%) urgent surgery. The incidence of postoperative AF was 27.6% in those patients who underwent isolated CABG and 41.0% in those who underwent CABG with concomitant valve surgery. Overall, 89.7% of the surgeries were performed on-pump. Patients with longer perfusion and cross-clamp times had a higher frequency of AF.
|Variable||Postoperative AF||p Value|
|No (n = 1,531)||Yes (n = 646)|
|Urgent surgery||748 (49%)||311 (48%)||0.7503|
|Isolated CABG||1,235 (81%)||471 (73%)|
|CABG and valve surgery||214 (14%)||149 (23%)|
|CABG with AVR||129 (8%)||106 (16%)|
|CABG with MV replacement||33 (2%)||20 (3%)|
|CABG with MV repair||67 (4%)||42 (7%)|
|CABG and other surgery||82 (5%)||26 (4%)|
|On-pump||1,371 (90%)||581 (90%)||0.7855|
|Cardiopulmonary bypass time||105 (81–136)||116 (88–148)||<0.0001|
|Cross clamp time||71 (53–99)||82 (59–113)||<0.0001|
We identified several factors that were associated with the development of postoperative AF ( Table 3 ). Increasing age, increasing height, body mass index >35, NYHA class IV heart failure, preoperative dialysis, and concomitant aortic valve replacement were all associated with a higher frequency of postoperative AF. Alternatively, a previous diagnosis of dyslipidemia and African American race were associated with a lower frequency of postoperative AF.
|Variables||OR||95% CI||p Value|
|Age (increment by 10)||1.74||1.53–1.97||<0.0001|
|Height in cm (increment by 10)||1.17||1.06–1.29||0.0020|
|Body mass index >35||1.64||1.24–2.18||0.0006|
|NYHA class IV heart failure||1.41||1.05–1.89||0.0205|
|Concomitant aortic valve replacement||1.63||1.23–2.18||0.0008|
Overall, 84% of the cohort was taking preoperative AF prophylaxis. As shown in Tables 4 and 5 , beta-adrenergic blocking agents were the most commonly used medications for prophylaxis (72%). Membrane active medications were used in 19%: 11% amiodarone, 7% other antiarrhythmic therapy, and 1% sotalol. The most commonly used combinations included beta blockade with calcium antagonist (9%) and amiodarone with either beta blockade or calcium antagonist (8%; Table 4 ). African Americans received prophylaxis at a slightly lower rate (78% vs 84% overall). Among African Americans, 72% received beta blockers (n = 64/89). Figure 1 shows the use of preoperative AF prophylaxis at the site-level for any medication, beta blockade, sotalol, and amiodarone. There was evidence of variability across sites in the use of any prophylaxis medication, beta blocker, and amiodarone.
|n = 2,177|
|No therapy||391 (18.0%)|
|BB only||1,100 (50.5%)|
|CA only||103 (4.7%)|
|BB + CCB||204 (9.4%)|
|Amiodarone only||45 (2.1%)|
|Amiodarone + AVN blocker(s)||168 (7.7%)|
|Other AAD||38 (1.8%)|
|Other AAD + AVN blocker(s)||90 (4.1%)|
|Sotalol only||6 (0.3%)|
|Sotalol + AVN blocker(s)||4 (0.9%)|
|All other combinations ∗||28 (1.3%)|