Non-valvular new onset supraventricular arrhythmia (SVA; atrial fibrillation, atrial flutter, atrial tachysystolia) is the most frequent rhythm disorder in surgical and non-surgical intensive care units (ICUs), occurring in 4.5% to 46% of critically ill patients . Although SVA in critically ill patients is now recognized as an indicator of illness severity, its independent relationship with mortality is the subject of debate . Regarding morbidity, recent data show that in-hospital arterial thromboembolic events (ATEs) are not rare . A multicentre retrospective study, with approximately 50,000 patients hospitalized for severe sepsis, reported that SVA was associated with an increased adjusted risk of in-hospital ischaemic stroke (2.6% vs 0.6%; odds ratio 2.70 [2.05–3.57]; P < 0.001) .

In cardiology patients, although numerous clinical trials have provided an extensive evidence base for the use of antithrombotic therapy to prevent SVA-related ATEs, bleeding risk limits its administration . Antithrombotic therapy to prevent ATEs is recommended by the European Society of Cardiology experts, in patients at intermediate risk or high risk according to the CHADS ₂ (Cardiac failure, hypertension, age ≥ 75 years, diabetes, stroke [doubled]) and, more recently, the CHA ₂ DS ₂ -VASc (Congestive heart failure, hypertension, age ≥ 75 years (doubled), diabetes, stroke [doubled], vascular disease, age 65–74 years, sex category [female]) risk stratification scores . In 2001, Gage et al., using data from the National Registry of Atrial Fibrillation, validated the CHADS ₂ score for predicting 1–3-year stroke rate in cardiology patients aged 65–95 years with non-valvular atrial fibrillation and no antithrombotic therapy at discharge . To improve risk stratification, Lip et al. extended the CHADS ₂ schema in 2010, by considering additional stroke risk factors (vascular disease, female sex and age ≥ 65 years) in the CHA ₂ DS ₂ -VASc risk score, validated to predict 1-year stroke rate in cardiology patients with non-valvular atrial fibrillation and no vitamin K antagonist or heparin treatment at discharge . To balance the risk of stroke against the risk of major bleeding, which is the most feared complication of anticoagulation therapy, a formal bleeding risk assessment with the HAS-BLED (hypertension, abnormal renal/liver function, stroke, bleeding history or predisposition, labile international normalized ratio, elderly [age > 65 years], drugs/alcohol concomitantly) score is recommended for all patients with atrial fibrillation .

Recently, the American Association for Thoracic Surgery developed guidelines for the prevention and treatment of perioperative/postoperative atrial fibrillation for thoracic surgical procedures . Anticoagulation within the first 48 h of new onset atrial fibrillation should be considered based on the patient’s CHA ₂ DS ₂ -VASc risk score (S) for stroke, weighed against the risk of postoperative bleeding (S = 0: no anticoagulation recommended; S = 1: anticoagulation should be considered if its benefits outweigh the risk of bleeding; S = 2: anticoagulation is highly recommended if its benefits outweigh the risk of bleeding; class I C recommendation). However, this suggested perioperative/postoperative thromboembolic risk stratification is based on indirect evidence from studies outside of the perioperative setting, involving patients with chronic non-valvular atrial fibrillation and no antithrombotic therapy . Moreover, no scoring system is recommended by these guidelines for identifying the risk of bleeding after initiation of anticoagulation .

In surgical and non-surgical critically ill patients with SVA, excluding patients in the postoperative stage of cardiac surgery, a single-centre prospective study by Champion et al. evaluated for the first time the CHADS ₂ and CHA ₂ DS-VASc risk scores in the prediction of SVA-related ATEs . Champion et al. reported 16 in-hospital SVA-related ATEs in 12 of 108 (11%) patients, including five strokes, three mesenteric infarctions, two limb ischaemias, two embolic myocardial infarctions and four left atrial appendage (LAA) thrombosis. The CHADS ₂ and CHA ₂ DS ₂ -VASc scores, previous SVA, mitral stenosis and the patient’s history of stroke were significantly related to the onset of ATEs. The most accurate threshold for predicting an ATE was a CHADS ₂ score of ≥ 4. Using a multivariate analysis, the patient’s history of stroke was the only independent predictive factor for ATEs in SVA patients . We cannot, however, conclude that the CHADS ₂ risk score can predict SVA-related ATEs in critically ill patients on the basis of this single-centre study. First, risk scores were validated in a homogeneous cardiology population with non-valvular atrial fibrillation and no antithrombotic therapy. Champion et al. attempted to validate these scores in a different heterogeneous population that included valvular SVA and patients with antithrombotic therapy . Second, the authors report a higher incidence of SVA-related ATEs (11%) than the incidence described previously in ICUs . The inclusion of patients with valvular SVA (mitral stenosis) and those with a history of SVA may explain these results in part. Third, the main outcome variable was onset of an ATE, including clinical and echocardiographic (LAA thrombus detected by echocardiography) events, even though LAA thrombus itself is a major risk factor for clinical ATEs. Fourth, the decision to give anticoagulants cannot be based on a CHADS ₂ score ≥ 4, given its low sensitivity (50%) and low positive predictive value (26%) in predicting SVA-related ATEs in ICU patients, notably because of their increased bleeding risk .

Given that neither risk factor nor risk stratification strategy is validated in the ICU setting, the assessment of the SVA-related ATE risk in critically ill SVA patients, particularly in the perioperative phase of thoracic surgery, is based on studies in a cardiology population . Recommendations for the prevention of ATEs with long-term antithrombotic therapy in cardiology patients are based on the presence (or absence) of clinical and echocardiographic risk factors for ATE with CHADS ₂ and CHA ₂ DS ₂ -VASc risk stratification scores. In particular, recent European Society of Cardiology guidelines recommend an anticoagulation strategy before elective cardioversion (spontaneous, electrical or pharmacological) of SVA in patients with stable haemodynamics: the strategy depends on whether the duration of the SVA is < 48 h or > 48 h . Since the 1990s, several observational transoesophageal echocardiography (TOE) studies have demonstrated LAA thrombus in 0% to 27% of patients with SVA of > 48 h duration . Although LAA thrombi have been identified by TOE within shorter intervals , recommendations are based on the assumption that a LAA thrombus requires a minimum of 48 h to form after the onset of SVA. Because of the strong correlation with pericardioversion ATEs, a thrombus in the left atrium or LAA detection with TOE prohibit cardioversion and require anticoagulation for at least 3 weeks .

These results from cardiology patients still remain to be demonstrated in surgical and non-surgical critically ill patients. In fact, the frequency of and the required time for LAA thrombus formation in critically ill patients with SVA are possibly different in the case of cardiology patients. Champion et al. reported a high incidence of ATEs and described an LAA thrombus in four of 108 patients (3.7%), although the duration of SVA was systematically < 48 h . Moreover, the incidence of LAA thrombus is possibly underestimated, as TOE was not systematically performed; transthoracic echocardiography without TOE leads to LAA thrombus detection with very low sensitivity. Therefore, these data could suggest a incidence of in-hospital ATEs (11%) that is higher than that observed in cardiology patients (2.3% stroke rate at 1 year ), despite the SVA duration being usually < 48 h . Intracardiac thrombus formation in cardiology patients with SVA begins with Virchow’s triad of stasis, endothelial dysfunction and a hypercoagulable state. Exacerbation of these pathophysiological mechanisms in critically ill patients presenting severe inflammation or sepsis could explain the results reported by Champion’s group. So, preventing the SVA-related ATE risk that occurs in ICUs requires specific studies for this very specific population.

As we are unaware of any studies reporting the frequency and required time for LAA thrombus formation in critically ill patients with SVA, the anticoagulation strategy in this context is not consensual. Thus, some studies reported that therapeutic anticoagulation is prescribed for 16% to 58% of critically ill patients with SVA . As these patients who received systemic anticoagulation had a high frequency of major bleeding events (9% to 45% depending on the study ), we recommend administering anticoagulation only in patients with the highest ATE risk.

To identify SVA-related ATE risk factors in critically ill patients, we have begun a TOE multicentre prospective observational study (N ^o ID-RCB: 2014-A00463-44). The objective of this study is to determine the clinical and echocardiographic risk factors for ATEs and, particularly, to determine the frequency of and the required time for LAA thrombus formation in patients with new onset SVA in ICUs.