A wolf in sheep’s skin? Postoperative atrial fibrillation after cardiac surgery and the risk of stroke and mortality

Postoperative Atrial Fibrillation (poAF; AF) is the most common complication after heart surgery.¹ Up to 50% of patients who present preoperatively in normal sinus rhythm will develop poAF after cardiac surgery making it the most common postoperative morbidity.² It is hypothesized that inflammation, increased sympathoadrenal activation, and oxidative stress unmask the propensity for developing poAF in at-risk patients, thereby leading to blood stasis and clot formation in the left atrium.³ Resulting thromboembolism and cerebrovascular accidents can dramatically change the postoperative course and affect patients’ lives. Despite multiple studies associating poAF with a 5-fold increased risk of developing long-term AF, prolonged ICU and hospital length of stay, stroke, and overall mortality,^{1, 4} other studies suggest poAF is a benign self-limiting disease.^{5, 6} Furthermore, while some studies report poAF to be an independent risk factor for neurologic events after cardiac surgery,^{7, 8} others did not make this association.^{9, 10}

In this issue of the Journal of Cardiothoracic and Vascular Anesthesia, Wang et. al. reported the results of a prospective cohort study assessing the relationship between poAF and stroke and mortality after cardiac surgery.¹¹ This study is based on 3,008 adult patient who underwent coronary artery bypass grafting or aortic valve replacement surgery. The authors based their study on a comprehensive prospectively collected database of adult cardiac surgery patients from two large tertiary care centers in Singapore. Patients were recruited between August 2008 and July 2012. According to the study protocol, demographic variables as well as perioperative risk factors and outcomes data were prospectively collected. Four patient subgroups were established: patients with no preoperative or postoperative AF, patients with pre-existing AF and poAF, patients with pre-existing AF but no poAF, and those with no pre-existing AF but with AF after surgery. To identify AF, patients were monitored via continuous telemetry for 72 hours postoperatively and then electrocardiograms (ECG) were performed daily or as needed based on patient symptoms. PoAF was defined as AF lasting for at least 1 hour and associated with hemodynamic instability or requiring medical treatment. All patients were followed up for postoperative stroke until December 2016 and mortality until June 2018. Stroke was defined based on clinical signs and symptoms suggestive of a subarachnoid hemorrhage, intracerebral hemorrhage, or cerebral ischemic infarction lasting for more than 24 hours without apparent non-vascular causes. Wang et. al. discovered that the incidence stroke was greater in patients with history of preoperative AF and in patients with a new onset poAF compared to those with no atrial fibrillation.¹¹ Furthermore, new-onset poAF was identified as an independent risk factor for postoperative stroke and was associated with increased mortality after cardiac surgery.

The undoubtful strength of this study is the prospective nature of the collected patient database. In general, the difficulty of poAF research originate from obstacles to accurately capture this rhythm disturbance during the postoperative period in all patients. The peak incidence of poAF occurs between the second and fourth postoperative day.¹² This poses challenges for accurate detection of AF as many patients are transferred out of the intensive care unit and are not monitored by telemetry at that time. This is further complicated by a transient, albeit often recurrent nature of poAF,^{1, 3} deficiencies in record keeping, difficulties in data extraction in the absence of electronic medical record systems during retrospective analyses, and differing definitions of AF in terms of duration and frequency. The prospectively collected database, a priori defined goals and objectives, meticulous data storage and patient follow up alleviated these challenges and improved the validity of the results. The relatively large patient sample size further strengthens the results of this study.

Wang et. al. utilized The Singapore Stroke Registry to identify the patients who suffered from a stroke during the follow up period.¹¹ While this database does not include patients who received care through the substantial private healthcare sector in Singapore, it has been reported to effectively capture stroke outcomes in the majority of cases.¹² In fact, the universal healthcare system implemented in Singapore in 1983 likely played a positive role in effectively capturing the outcomes.

This study expands our knowledge of the morbidity and mortality associated with poAF and helps us better appreciate the significance of the problem. The impact of poAF and its sequelae on the postoperative course and quality of life of our patients is detrimental. The revealed associations between poAF and increased risk of stroke and mortality underscore the importance of identification of high-risk patients and timely and targeted prophylaxis. Current poAF prevention efforts are directed at development of accurate prediction models based on risk factors. One of the risk factors most widely reported in the literature is age.^{1, 13} The most recent practice advisory by the Society of Cardiovascular Anesthesiologists/European Association of Cardiothoracic Anaesthetists, also published in this journal, identifies age as a significant risk factor for poAF after cardiac surgery and suggests additional poAF prophylaxis for patients > 75 years old.^{14, 15} Interestingly, Wang et. al. do not support the association between age and increased risk for poAF.¹¹ This may be explained by the relative youth of patients in the study (mean age 59.3) and the overall health status of the population in Singapore, whose citizens enjoy one of the longest life expectancy and healthy life expectancy in the world.

White race is another risk factor reported to be associated with an increased risk of poAF.¹ Whilst no data are presented for the at-risk category ‘white ethnicity’ in the sample, this study makes an important contribution to the body of literature by specifically categorizing Chinese, Indian, and Malay patients. These ethnic groups are often not described separately in the studies originating from North America and Europe. Genetic predisposition is a plausible explanation for racial differences in poAF risk. In fact, common AF has a known substantial genetic component, evidenced by linkage analyses and genome-wide association studies. Even though this study did not detect poAF risk differences between ethnical groups, it was not specifically powered for this goal. Genotypical heterogeneity of poAF can be a promising avenue for future research to help identify high risk patients and personalize prophylactic measures. International genetic databases obtained from diverse populations are of particular interest for this purpose.

While Wang et al. did not differentiate between thromboembolic and hemorrhagic stroke outcomes in their study, timely anticoagulation in poAF patients is to be balanced against the inherent risks.¹¹ The anticoagulation strategies suggested to prevent thromboembolic phenomena and stroke are in fact a double-edge sword as they may result in hemorrhagic complications, including stroke. Unfortunately, anticoagulation medication data were not available in this study and should be included in future prospective investigations.

The wide spectrum of parameters that may increase risk of poAF after open heart surgery, including patient demographics and comorbidities, as well as surgery-related risk factors complicates study design; accounting for all the co-variates is extremely challenging. This inherent bias of cohort studies can play a particular role when non-diverse patient populations are being sampled. Therefore, the external validity of the results may be affected. In the future, machine learning approaches may be an attractive additional tool to perform data analysis. This emerging technology allows for analysis of large patient datasets via algorithms utterly different from conventional logistic regression methods and helps adjusting for multiple covariates. Future multi-institutional large prospective trials, with continuous telemetry during the entire poAF high-risk postoperative period and integration of machine learning and conventional statistical approaches may further help to provide convincing evidence that poAF is not a transient postoperative phenomenon, but a morbid condition that leads to ominous complications and increases mortality.