Download PDF

Atrial fibrillation (AF) is a common complication of coronary artery bypass surgery (CABG), with an incidence of 20% to 40%.1 It usually occurs within the first week, with a peak incidence on the second postoperative day.2 Technical advances in cardiac surgical and anesthetic techniques have not significantly decreased the prevalence of post-CABG AF, and there is some evidence that it may be increasing due to the increasing age of the population undergoing revascularization.3

Atrial fibrillation after CABG was thought to be a benign, self-limiting condition, but more recent evidence suggests that postoperative AF is independently associated with increased in-hospital stay and long-term mortality.4 Its occurrence is linked with early postoperative complications such as hemodynamic instability with inotropic support, prolonged ventilation or re-intubation, intra-aortic balloon pump use and reoperation for bleeding.5 Although infrequent, cerebrovascular events can be a serious and potentially life-threatening complication. 4,6 The development of AF, even without haemodynamic compromise, requires additional medical and nursing time and is associated with increased duration and cost of hospitalisation.3

Pericardial collections constitute a recognised complication following CABG. 7 They can range from minimal pericardial effusions with no direct haemodynamic effects to bigger collections that can cause mechanical compromise and tamponade requiring emergency intervention. Previous reports show a relationship between the development of pericardial effusions and AF post CABG. 7,8 A recent meta-analysis of 6 prospective randomised studies has shown an overall reduction in pericardial effusions with the creation of a posterior pericardiotomy.9 We hypothesised that prevention of pericardial collection by creating a pericardial window at the time of operation may favourably affect AF incidence after CABG. We also implemented a novel approach of not placing a mediastinal or pericardial drain and to rely only on the posterior pericardial window to drain the mediastinum into the left pleura, where a pleural chest drain was routinely inserted. We set out to examine this practice in a continuous series of patients undergoing elective CABG.

Materials and Methods

Data was prospectively collected from 108 consecutive patients undergoing elective CABG under the care of one surgeon (JD) at King’s College Hospital (London, UK) between May 2001 and November 2002. With ethical and institutional approvals, informed consent was obtained from all patients. The first 50 consecutive patients were allocated to the study group (posterior pericardial window with pleural drain and no mediastinal drains) and the following 58 consecutive patients were allocated to the control group (no pericardial window with placement of mediastinal drains). Eligible patients included those for first-time, elective CABG with no history of AF or other arrhythmias, or antiarrhythmic use. Presence of permanent pacemaker, valvular heart disease, acute coronary syndrome within the previous month and renal dysfunction were excluded. In addition, patients with inflammatory diseases that might lead to chronic pericarditis were also excluded.

Standard anaesthetic and perfusion protocol were used. Briefly, anaesthesia was maintained using a mixture of propofol and remifentanil. Blood gas analysis and activated clotting time (ACT) were checked regularly tomaintain a minimum ACT of 400 seconds during cardiopulmonary bypass. Patients were cooled to 32°C. A cardiac index of 1.8 – 2.4 L/m2/min was used to determine each patient’s target normothermic cardiac output. Myocardial protection was achieved using cold (4:1 ratio) blood-cardioplegia. The study group had a posterior pericardial window created just prior to discontinuation of cardiopulmonary bypass. The heart was gently retracted to access the posterior pericardium and a 4 cm cruciate incision made parallel and posterior to the left phrenic nerve using diathermy. No mediastinal chest drain was used and a pleural chest drain was placed according to the number of internal mammary artery harvests or pleural openings. The control group without a pericardial window had one anterior mediastinal chest drain (no drains behind the heart) and pleural drain as in the study group.

Chest drains were placed on continuous suction (-5 kPA) and removed on postoperative day 1 as per protocol, in the absence of significant mediastinal bleeding, pneumothorax and anything but minimal drain (<30 mL) output for two consecutive hours.

Postoperatively standard unit protocols were used. Briefly both groups received prophylactic β-blocker from day one postoperatively unless contraindicated due to persistent hypotension or bradycardia. Chest infection was diagnosed on the basis of strict guidelines including clinical (purulent sputum, elevated temperature), elevated white cell count, pathological (sputum samples; microbial culture and sensitivity test) and radiological signs.

All subjects underwent two-dimensional transthoracic echocardiography by the same operator (blinded to treatment) four days postoperatively, to evaluate left ventricular function, left atrial dimensions and presence of pericardial effusion. Any pericardial effusion >1cm was considered significant.

Patients were telemetrically monitored for 96 hours to detect AF or any supraventricular tachycardia, which was confirmed by standard 12-lead electrocardiogram. A persistent arrhythmia (greater than 30 minutes) was treated with correction of serum potassium to greater than 4.5 mmol/l and 20 mmol/l of magnesium if required. If the arrhythmia persisted patients were commenced on amiodarone. Patients were followed until discharge and heart rhythm was assessed with daily electrocardiograms and 6-hourly clinical examinations for the duration of their hospital stay (beyond the first four days of continuous telemetry). All data was prospectively collected.

Statistical analysis:

Continuous data was expressed using the mean ± standard deviation. Baseline and outcome variables between the window and non-window group were compared using Pearson’s chi-square test or Fisher’s exact test for dichotomous data, whereas unpaired t-test was used for continuous variables with normal distribution. Threshold for rejecting null hypothesis (a-error) was set to p-value 0.005 which corresponds to p-value 0.05 after adjusting for multiple testing. The analyses were conducted using Stata version 9.1 (Statacorp, College Station, Texas).


All patients initially allocated to the study and control groups continued in the trial and were included in the analysis. There were no significant differences in age, diabetes, hypertension, Canadian Cardiovascular Society Score, previous myocardial infarction and left ventricular function between the two groups (Table 01). Equally, intraoperative data was not significantly different in terms of number of distal coronary anastomoses, mean aortic cross-clamp and cardiopulmonary bypass times between the two groups (Table 02). With regards to postoperative outcomes, there was no significant difference in postoperative total blood loss, intubation time, pleural effusions and length of hospital stay. Finally, there was no need for re-exploration for bleeding in our patient cohort.

There were two deaths, one in the control and one in the study group. Both exhibited low cardiac output syndrome and suffered VF arrest on days 6 and 20, respectively.

In the control group 9 patients (15.5%) developed significant pericardial effusions (>1 cm) compared with none (0%) in the pericardial window group (p = 0.004). Twelve patients in the control group (21%) compared with three patients (6%) in the pericardial window group developed postoperative AF (OR : 0.24, 95% confidence interval: 0.06 to 0.92, p = 0.038).

The pericardial window group had significantly less blood-transfused post operatively (224 vs. 621 mL, p < 0.001). However, the pericardial window group exhibited a significantly increased rate of chest infection compared to control group (22% vs 4%, p = 0.003). There was no difference in length of hospital stay as a result of pericardial window compared with controls, i.e., 10.8 vs. 9.7 days (p= 0.36), respectively).

Comparison of baseline characteristics between the study and the control groups.

Variable Pericardial windowN = 50 n=49 ControlN = 58 p value
Age (years) 65.7 ± 9.0 67.2 ± 8.7 0.38
Male 43 (86%) 43 (74.1%)
BMI 27.3 ± 4.2 26.9 ± 3.6 0.6
Non-diabetic 36 (72.0%) 45 (77.6%)
Diet controlled 2 (4%) 1 (1.7%) 0.86
Tablet controlled 5 (10%) 5 (8.6%)
Insulin 7 (14.0%) 7 (12.07%)
Pulmonary disease
None 43 (86.0%) 49 (84.5%)
Asthma 1 (2.0%) 1 (1.7%)
Emphysema 1 (2.0%) 3 (5.2%) 0.51
COPD 3 (6.0%) 5 (8.6%)
Previous TB 2 (4.0%) 0 (0%)
Smoking History
Non-smoker 12 (24.0%) 17 (29.3%)
Ex smoker >5 years 20 (40.0%) 20 (34.5%) 0.80
Ex smoker <5 years 10 (20.0%) 14 (24.1%)
Smoker 8 (16.0%) 7 (12.1%)
Hypertension 34 (68.0%) 38 (65.5%) 0.79
Angina classification (CCS)
None 4 (8.0%) 8 (13.8%)
Class 1 4 (8.0%) 3 (5.2%)
Class 2 23 (46.0%) 15 (25.9%) 0.08
Class 3 18 (36.0%) 25 (43.1%)
Class 4 1 (2%) 7 (12.1%)
Severity of CAD
Left main stem 3 (6.0%) 6 (10.3%) 0.27
Three vessel disease 39 (78.0%) 37 (63.8%)
Previous MI 24 (48.0%) 29 (50.0%) 0.84
LV function
Good (≥ 50%) 29 (58 %) 33 (57%)
Fair (30% ≥ and < 50%) 16 (32%) 16 (28%) 0.67
Poor (< 30%) 5 (10%) 9 (16%)

Note. Categorical data are expressed as numbers (%); continuous data as means ± standard deviation. BMI: body mass index; COPD: chronic obstructive pulmonary disease; TB: tuberculosis; CCS: Canadian Cardiovascular Society; CAD: coronary artery disease; MI: myocardial infarction; LV: left ventricular.

Table. 02

Comparison of intraoperative variables and postoperative outcomes between the study and the control groups

Variable Pericardial window Control p value
Number of grafts 3.2 ± 0.1 2.9 ± 1.0 0.09
Cross clamp time (min) 40.8 ± 29.2 49.2 ± 29.5 0.15
Cardiopulmonary bypass time (min) 62.7 ± 47.5 84.4 ± 43.0 0.016
Intubation time (hr) 6.0 ± 4.8 5.8 ± 7.1 0.85
Total blood loss (mL) 898 ± 503 962 ± 512 0.53
Chest infection 11 (22.0%) 2 (3.5%) 0.003
Length of hospitalstay (day) 10.7 ± 6.8 9.7 ± 5.0 0.36
Pericardial effusion (> 1 cm) 0 (0%) 9 (15.5%) 0.003
Postoperative AF 3 (6.0%) 12 (20.7%) 0.003

Note. Categorical data are expressed as numbers (%); continuous data as means ± standard deviation.


The pathogenesis of AF after cardiac surgery appears to be multifactorial. Mechanical factors as well as technical parameters contribute to the already recognised effects of inflammation, oxidative stress and pre-existing susceptibility of the atrial substrate.1 Consistent clinical predictors of AF after CABG are advanced age and left atrial enlargement. 1,2 Induction of any cardiac arrhythmia requires the combination of an appropriate trigger in the presence of a susceptible substrate. Structural remodeling with atrial fibrosis and apoptosis, increased oxidative stress, and metabolic modifications leading to an energy demand-supply mismatch, seem to contribute and precede the development of postoperative AF.10

Numerous studies suggest a beneficial role of β-blocker use in patients undergoing cardiac surgery, which is currently considered routine practice.11 Concurrently, several studies have demonstrated satisfactory efficacy of known antiarrhythmics, such as amiodarone and sotalol, in the prevention of postoperative AF.12 The narrow therapeutic spectrum, proarrhythmic potential and increased toxicity of these agents hinder their routine application for postoperative AF prophylaxis. The emerging concept of “upstream” therapies (statins, polyunsaturated fatty acids, angiotensin-converting-enzyme inhibitors / Angiotensin II receptor antagonists and antioxidant vitamins) which aim mainly at the modulation of the susceptible substrate has shown promising results in postoperative AF prophylaxis, avoiding the side-effects of conventional antiarrhythmics.13

Pericardial collections are some of the established and possibly avoidable triggers for postoperative AF.7 The exact mechanism for this association can only be speculated; tissue ischaemia due to diastolic compromise as well as the possible local release of inflammatory mediators from haemorrhagic effusions could lead to an energy supply-demand mismatch that has been previously implicated in the pathogenesis of postoperative AF.10,14

Previous studies have highlighted the potential role of posterior pericardiotomy in the prevention of postoperative AF 8,1519 and a more recent meta-analysis incorporating these studies has further supported this preventative modality.9 It is suggested that the beneficial effect of posterior pericardiotomy on AF is driven by the more efficient drainage of possible postoperative effusions. 8,17 However, the routine use of mediastinal drains and their variable positioning could also influence the isolated effect of posterior pericardiotomy in mediastinal drainage. In our study we elected not to use mediastinal drains in our pericardial window group, in the belief that this may result in an additional benefit on dysrhythmias due to the lack of direct contact of the silicon tube with the heart, and an effect on patient comfort.

Despite the non-randomised nature of the study, there were no significant differences in baseline characteristics and intraoperative variables between the posterior pericardiotomy and control groups. The beneficial role of posterior pericardiotomy in reducing new-onset of AF was associated with the absence of pericardial effusion despite the non-utilisation of mediastinal drains. The total blood loss was similar in the two groups, suggesting that the additional pericardial intervention does not contribute to bleeding as previously reported.16 It is important, however, to mention the significantly higher rate of chest infections in the pericardiotomy group, an observation recently raised.9 The absence of mediastinal drain led to increased load to the often-used single pleural drain, leading to prolonged accumulation of fluid within the chest cavity, causing atelectasis and infection. This complication, however, did not lead to significantly increased hospital stay in the intervention group, as this may have been counterbalanced by the significantly reduced rate of AF, which is known to prolong discharge from hospital.

Study limitations:

Due to the non-randomised nature of this study, and despite the homogeneity of the treatment and control groups in baseline characteristics, there is a possibility for selection bias that we could not otherwise control for. In addition, although the peak incidence of AF occurs on postoperative days 2 and 3 it is likely that short, asymptomatic episodes of AF may have occurred beyond day 4, when tele-monitoring was discontinued, and which may have not been captured with daily ECGs and 6 hourly clinical examinations. Despite the observation of a significant beneficial effect of pericardiotomy on postoperative AF, the mechanism can only be speculated; its identification was beyond the scope of this study.


A posterior pericardial window with pleural drainage may avoid the need for mediastinal drains and reduce the incidence of postoperative pericardial effusion and atrial fibrillation in patients undergoing coronary artery bypass surgery.


This research was supported by the NIHR Bristol Cardiovascular Biomedical Research Unit.