Secondary bacterial infections trigger New Onset Atrial Fibrillation in ICU Covid-19 ARDS patients

Cardiac arrhythmias, mainly atrial brillation (AF), is frequently reported in COVID-19 patients, yet causality has not been explored. Intensive Care Unit patients frequently present AF during critical illness. Sepsis is one of the main contributors of AF occurrence in ICU patients. The aim of the study was to explore if Covid-19 myocardial involvement is the only contributor for New Onset Atrial Fibrillation (NOAF) in intubated ICU patients.

Therefore, as respiratory worsening occurs after several days from symptom onset, AF seems to occur late in the course of the disease, when the patients have been admitted in the ICU [1,13,14]. Severe Covid-19 could be a risk factor, yet, in ICU, other-than SARS-CoV-2 trigger factors may be implicated, which have not yet been explored.
In the non-Covid-19 era, NOAF is a frequently encountered arrhythmia in ICU patients [16]. Multiple factors are implicated, leading to structural or electrical remodeling of the atria, triggering AF, sepsis being the leading one [16]. Sepsis could be a possible trigger factor in Covid-19 as well, as the incidence of ICUacquired secondary infections is increased, partly attributed to the immunosuppressive drug protocols adopted [18][19][20][21].
In the present study we hypothesized that other factors than SARS-CoV-2 infection may contribute to NOAF occurrence in ICU patients with Covid-19 ARDS. Therefore, we aimed to evaluate in a cohort of intubated, ICU, Covid-19 ARDS patients: 1). the incidence of NOAF and investigate possible factors leading to its occurrence, 2). the course of NOAF during ICU stay, and 3). the cardiac involvement using echocardiography and troponin levels upon ICU admission.

Methods
Study population. Consecutive patients admitted in the ICU (March 2020-February 2021) at the University Hospital of Larissa, Greece with laboratory con rmed (real-time PCR) SARS-CoV-2 infection and ARDS were included in this prospective study. The study was approved by the local ethics committee (55951/2020), with a waiver for informed consent. All patients were admitted intubated, and evaluation included the period from the rst ICU day until the 28th day (either still in the ICU, discharged or dead).
All eligible (according to inclusion/exclusion criteria) patients were divided in two groups: the NOAF group including patients with new onset AF during ICU stay without previous history of AF, and the control group, including all other patients not presenting AF. Patients were included in the NOAF group if they presented at least one AF episode lasting more than 10 minutes or suffered multiple AF episodes during a 24-hour period, or AF episodes needing direct electrical cardioversion due to hemodynamic instability. Exclusion criteria were: 1). history of recent myocardial infarction or previous echocardiography presenting wall motion abnormalities indicating ischemic disease, 2). recent admission for either coronary artery bypass graft, cardiac surgery or percutaneous transluminal coronary angioplasty (PTCA), 3). severe aortic or mitral stenosis or regurgitation, 4). patients with a history of heart failure from any cause or previous echocardiographic ndings indicating Left Ventricular Ejection Fraction (LVEF) below 45%, 5). known Right Ventricular (RV) dysfunction, 6). cardiomyopathy of any type, 7). presence of pacemaker, 8). congenital heart disease, 9). brief AF episodes not meeting inclusion criteria, 10). history of NOAF occurring in the ward or presenting upon ICU admission, 11). death during the rst 48 hours of ICU admission, 12). permanent AF, 13) history of paroxysmal atrial brillation (PAF). However, PAF patients were analyzed separately (Appendix).
Data collection. Baseline characteristics and disease severity (APACHE II, SOFA score) were recorded on admission. Demographics, medical history, and data concerning COVID-19 infection prior to hospital admission were collected from patients' medical records and/or next-of-kin. Laboratory ndings [in ammation markers (CRP, ferritin), coagulation, electrolytes] and the SOFA score were recorded daily.
Troponin levels were recorded on ICU admission, the NOAF day and whenever indicated, according to attending physicians.
Blood, urinary and endotracheal aspirate (ETA) cultures were collected on admission, every 3 days (per local protocol) and whenever indicated, according to the attending physicians, but always on the day of AF occurrence and the day after. Heart rhythm was assessed continuously from the patients' monitor (General Electric, Carescape B850); ECG tracings (12-lead) could be reviewed for the preceding 72-hours (GE monitor's software), while 12-lead ECG was performed daily.
All patients received enhanced prophylactic anticoagulation according to current suggestions for Covid-19, except in patients with contraindication (coagulation abnormalities, thrombocytopenia, active bleeding).
Echocardiography was performed on admission and whenever indicated, but always on the AF day, according to AHA guidelines [22], by trained operators (VT, NK, VV), (General Electric, Vivid E95). We used a standard procedure to assess LV and RV sizes, function and lling measurements (2D imaging, color doppler, Tissue Doppler Imaging (Appendix) [23]. Left ventricular myocardial performance was assessed using the two-dimensional speckle-tracking method [23].
In patients needing cardioversion after 24 hours due to AF persistence, a transesophageal echocardiography (TEE) preceded. If pulmonary embolism (PE) was suspected, Computed Tomography Pulmonary Angiography (CTPA) was performed.
All patients were assessed under satisfactory loading conditions (Appendix). Central Venous Oxygen Saturation (ScVO 2 ) measurements were performed on admission, in episodes of hemodynamic instability, and whenever indicated according to attending physicians.

De nitions
Sepsis, septic shock, and types of infections were de ned according to recently updated terms [24,25]. Secondary infections included all hospital acquired Blood Stream Infections (BSI), Hospital/Ventilator-Associated Pneumonia (HAP/VAP) and Urinary Tract Infections (UTI) occurring after 48 hours of hospital admission (Appendix).

Atrial brillation management protocol
All NOAF episodes received amiodarone (750 mg daily) after a loading dose of 150-300 mg administration (± b-blockers for rate control). Direct electrical cardioversion was performed only in patients with hemodynamic instability, de ned as a signi cant increase in vasopressor dosage after AF appearance, according to the attending physicians' assessment. However, attending physicians were encouraged to postpone electrical cardioversion until the patient had received the amiodarone loading dose.
Statistical analyses. Results are given as mean (± Standard Deviation, SD) in normally distributed parameters and as median (± Standard Deviation, SD) in not normally distributed values. Normally distributed continuous indices were compared with Student's t-test (between two groups) and one-way ANOVA (for multiple group and other repetitive variable measurement comparisons); non-normally distributed indices were compared via the Mann-Whitney-U and Wilcoxon test. Finally, Chi-square was used when testing categorical data. Data were analyzed using SPSS (IBM SPSS statistics version 25).
New Onset AF occurred late in the course of hospitalization, 18±4.8 days from Covid-19 symptoms' onset and on 8.5±2.1 ICU day. Only one patient (70-years-old) presented a short AF episode lasting 3 min (2 nd ICU day) and another two (65 and 72-years-old) presented with NOAF on ICU admission and were excluded. In all other patients, NOAF occurred after/on the 3 rd ICU day (range 3-23 ICU day).
Laboratory and clinical data on AF day, compared to data on the third preceding the AF day are presented in Table 2.

Septic secondary infection episodes
Twenty ve (41.6%) patients in the control group (upon ICU admission in two) and 17 (89.5%) in the NOAF group presented at least one secondary infection episode during their ICU stay (p<0.001) (Fig. 1).
Secondary infections occurred after the sixth ICU day in 30 (71.4%) patients.
NOAF group. Sixteen patients (84.2%) presented a septic secondary infection concurrently with NOAF, thirteen of whom (81.3%) presenting with septic shock. Eleven patients (57.9%) had primary bacteremia and ve (26.3%) suffered ventilator-associated pneumonia (VAP)-two accompanied with bacteremia. Pan-drug resistant (PDR) or extensively drug resistant (XDR) Gram (-) bacteria (Klebsiella pneumoniae and/or Acinetobacter baumannii) were the main isolates. All NOAF patients had negative blood, urine and ETA cultures on admission. In 14/16 patients, NOAF presented during the rst septic episode. Notably, antibiotics were added or modi ed in 18/19 patients, in the two preceding or the NOAF day, for suspected septic secondary infections, according to the attending physicians. Interestingly, in one of the two patients with NOAF upon ICU admission (excluded per protocol), the blood cultures drawn on admission revealed Acinetobacter baumannii.

Laboratory ndings
CRP values showed a gradual increase during the days preceding AF (Table 2) and a subsequent decrease thereafter.
Troponin levels signi cantly increased on the AF day compared to admission (0.64±1.04 vs 0.16±0.31 ng/dl, p=0.017). The highest value was observed in one patient (4.43 ng/dl) subsiding over the next days. This patient, with primary bacteremia presented laboratory and echocardiographic signs of septic cardiomyopathy.

Echocardiographic ndings
Left Ventricular function did not differ between groups, although RV function was mildly impaired in the control group. (Table 3). Left Ventricular Global Longitudinal Strain (GLSLV) was -12±4% in patients without AF and -14±6% in NOAF patients. Noteworthy, GLSLV was abnormal (<16.6%) in 78% of the patients in the control and 63% in the NOAF group [26]. Covid-19 patients presented moderate RV dilation (Right Ventricular End Diastolic Area/Left Ventricular End Diastolic Area>0.6 in both groups). Twenty-six (43%) vs ten (52%) in the control and NOAF groups respectively, presented mild pericardial effusion (p=0.48). Echocardiographic measurements did not signi cantly change on NOAF occurrence (Table 3).

Outcome.
Arrhythmia In all patients, NOAF lasted more than one hour. Sixteen patients (84%) returned to sinus rhythm (SR), 13 during the rst 24 hours and the rest during 48 hours, although short recurrent AF episodes (after cardioversion), lasting less than 30 minutes were recorded in 4 patients. Only one patient, presenting severe hemodynamic instability, was electrically cardioverted, one hour after unsuccessful amiodarone infusion; in this patient, AF recurred, returning to sinus rhythm after 24 hours. Nine TEE were performed in ve patients; in all, cardiac chambers, including the appendage were free from thrombi.
In six patients (including three in whom SR was not restored) signs of sepsis were not resolving until death. In three, AF recurred after 4-9 days on SR, coinciding with a new septic episode, returning to SR with sepsis resolution. Amiodarone infusion was continued until ICU discharge or death.
Mortality: In NOAF group, 28 th day mortality was 47% (9/19) vs 41.7% (25/60) in the control group (p=0.57). Three patients died without converting to SR. All ten survivors were discharged on SR, under amiodarone. Among them, we were able to contact three patients (aged 45, 56, 76 years) discharged home; patients are on SR (2-7 months later); amiodarone has been stopped.

Discussion
Our study demonstrates that myocardial dysfunction is present in intubated ICU patients with severe Covid-19 ARDS, as depicted by the echocardiographic ndings of impaired left and right ventricular function, mild pericardial effusion and mild elevation of troponin levels. However, New Onset Atrial Fibrillation occurred in ICU, Covid-19 ARDS patients, mainly in relation to a secondary infection that led to severe sepsis/septic shock. Demographics, ARDS severity, respiratory system mechanics, mechanical ventilatory modes and electrolytes did not differ between groups, while hypoxemia degree was quite improving in NOAF patients on the day AF occurred. We suggest that sepsis triggered NOAF occurrence, in the setting of an affected, from Covid-19, myocardium; sepsis resolution was crucial to maintain SR (under amiodarone infusion).
Our understanding on the cardiovascular effects of Covid-19 is still limited [4]. In our cohort, Left Ventricular EF was rather normal, although impaired global longitudinal strain indicated occult myocardial injury in the majority (74%) of Covid-19 ARDS patients upon ICU admission; abnormal GLSLV (< 16.6%) has been reported in 42% of Covid-19 patients admitted in the ward, while data on ICU patients are scarce [26][27][28][29]. In addition, a moderate RV enlargement was observed, which is in accordance with various Covid-19 reports [30,31]. However, multiple factors may explain this nding apart from Covid-19; RV dilation is exacerbated by mechanical ventilator settings (PEEP), especially when lung compliance is preserved [32,33]. Interestingly, 45.6% of the patients had a mild pericardial effusion. Pericardial effusion incidence has not been thoroughly evaluated in Covid-19 [34,35]. In addition, troponin levels were elevated on ICU admission, a nding that has been linked to myocardial involvement in Covid-19 [3][4][5][6]. The above data support the notion that a degree of myocardial injury is present in severe Covid-19 patients, admitted in the ICU [36]. However, there was no difference between the NOAF and control group, in any parameter concerning the cardiac involvement.
Among atrial arrhythmias, AF is the most frequent in Covid-19 patients; NOAF prevalence varies between 3-10% in non-ICU patients [10,11]. In our study, NOAF incidence was 24%, which is in accordance with the higher incidence reported in ICU patients [9-11, 13, 37-39]. Colon et al, noted a NOAF incidence of 16.5% in ICU patients [14]. However, no reference is made on possible secondary conditions and the timing of arrhythmia occurrence. Increased in ammatory markers and vasopressor need were reported during AF appearance, without specifying whether AF occurrence was coincidental to a secondary infection episode [2,9]. Other studies con rm the increased NOAF incidence in ICU Covid-19 patients, varying between 16.5-40%, yet, without specifying whether the virus or other factors, frequently present in critically-ill patients, are associated to its occurrence [9,10,11,[13][14][15][37][38][39]. Similarly, existing data lack information about the exact time of NOAF appearance in the course of Covid-19 [10,11,13,14,[37][38][39]]. An early, in the course of the infection, virally driven hyperin ammation-cytokine storm has been proposed as a possible mechanism, partially explaining NOAF occurrence in patients hospitalized in the wards [2,9]. In our study, NOAF appeared late in the course of the disease, approximately during the 18th post symptom onset day (8th ICU day), when COVID-19 symptoms usually subside [39]. Although myocarditis has also been suggested as a possible mechanism for arrythmias in Covid-19, histological ndings indicate macrophage in ltration, with no clear association to myocardial injury, and, although troponin is high, myocarditis is established in only 4.5% of the severely ill, Covid-19 patients with heart failure, undergoing endomyocardial biopsies; thus, the virus does not seem to directly invade the cardiac cells in order to initiate AF [12,40]. Our ndings support that non cardiac causes, such as systemic infection, may contribute to NOAF.
New-onset AF is a common arrhythmia in non-Covid-19 ICU patients, occurring in 19-35% of patients, sepsis being the main triggering factor [16, 41,42]. Walkey et. al, reported an increased incidence (35%) of NOAF among septic patients, further increasing with disease severity [16]. In our study, 84.2% of the patients presented sepsis and 68.4% had septic shock on NOAF episodes. In ammation markers, vasopressor need, and lactate levels presented a gradual increase in the preceding the AF days. The positive uid balance during the last three days, and the rise in ScVO 2 values, further supported NOAF's association to sepsis-induced vasodilation [43].
An increased incidence of secondary infections has been observed in our cohort; 42% in the control group and 89% in the NOAF group, consistent with recent reports. Buetti et al, reported an increased daily risk (HR 4.5) to acquire an ICU-BSI in Covid-19 compared to non-Covid-19 patients; BSIs usually occurred after the 7th ICU day [17]. Similarly, Rouze et al, reported that ventilator associated lower respiratory tract infections were more frequent in Covid-19 ARDS patients (50.5%) compared to in uenza (30.3%) and ICU patients with non-viral infections (25.3%), similarly occurring after the 7th ICU day [18]. Corticosteroids, Tocilizumab and Anakinra, used in COVID-19 ARDS, may be partly responsible [17,18,44]. In our study, NOAF occurred on the 8.5 ± 2.1 ICU day, mostly coinciding with the rst BSI/VAP septic episode. We suggest that sepsis, with adrenergic overstimulation, due to endogenous elevated catecholamine levels and exogenous catecholamine administration (as in septic shock), constitutes the second "hit", to trigger AF in a diseased/affected, from SARS-CoV-2, myocardium [2,16]. Interestingly, patients in both groups were of comparable age and did not present factors indicating apparent cardiovascular disease, known to increase AF occurrence risk [11].
Troponin levels were signi cantly raised on the AF day compared to admission, further supporting the association of NOAF to secondary sepsis. Troponin elevation has been repeatedly reported in bacterial sepsis, re ecting altered cardiomyocyte permeability or some degree of necrosis, frequently associated with cardiac dysfunction [45,46]. Sepsis induced myocardial dysfunction is very frequent, attributed to increased circulating catecholamine and cytokine levels, found in severe sepsis and septic shock [47,48].
However, decreased systemic vascular resistance may mask the altered myocardial performance. We believe that sepsis-induced vasoplegia is responsible for the apparently preserved LVEF in our patients when NOAF appeared.
Rhythm control has been found more bene cial than rate control in ICU patients [42]. Most patients in our study returned to sinus rhythm with pharmacologic cardioversion (amiodarone); AF could not be restored in patients with non-resolving sepsis or re-occurred in those with recurrent septic episodes.
Our study has limitations. It was conducted in a single center serving an urban population, thus the number of patients is limited. However, although our ndings may not be generalizable across the world, they may have particular importance in South Europe and other countries with an increased incidence of nosocomial infections from PDR/XDR, as in our study [49]. In addition, we consider an advantage that the study population was rather homogenous: we prospectively enrolled consecutive, intubated patients with severe Covid-19, with no obvious preexisting cardiovascular disease in order to eliminate known factors triggering AF. Cardiac Magnetic Resonance tomography was not performed, but its utility in ICU is limited by the out-of-hour availability and the requirement for some breath-holding, while no patients underwent endomyocardial biopsy (caring inherent risks), as it is not suggested due to the low incidence of myocarditis in Covid-19 [12]. Instead, in all patients, troponin levels and a full echocardiographic examination were performed, which seem appropriate to reveal cardiac involvement in Covid-19.

Conclusion
In conclusion, myocardial function is altered in Covid-19 patients, probably lowering the threshold for arrhythmogenicity. Secondary infections seem to be major contributors for NOAF occurrence in ICU Covid-19 ARDS patients, probably playing the role of the "second hit" in an affected myocardium. Sepsis should be suspected in case of late NOAF occurrence in these patients. Furthermore, AF did not resolve or re-occurred if sepsis persisted. Further research on the arrhythmogenicity of COVID-19 in severe ICU Covid-19 ARDS patients is needed.

Funding
This research did not receive any speci c grant from funding agencies in the public, commercial, or notfor-pro t sectors.

Availability of data and materials
The data set used for this manuscript will be available from the corresponding author upon reasonable request.
Ethics approval and consent to participate. This study was approved by the Institutional Ethics Committees of the University Hospital of Larissa (55951/2020), with a waiver for informed consent.

Consent for publication
Not applicable

Competing interests
The authors declare that they have no competing interests.  Positive ETA, N(%) 0 5 (26%) a For troponin and ScVO 2 , the value in the rst column refers on admission data b Eleven patients presented primary bacteremia on the day NOAF occurred. Two more patients diagnosed with VAP on NOAF day, presented positive blood cultures, with the same isolate as the one responsible for VAP (bacteremic VAP) Table 3. Comparison of echocardiographic variables between the control and NOAF group on admission, and between admission and the NOAF day, in the NOAF group.