Discussion
Our study including a single center cohort of 390 COVID-19 patients revealed 28/390 (7.2%) cases with documented cardiac arrhythmias, including 20 cases of AF, 3 atrial flutter, 1 VF (in a patient who also had paroxysmal AF), 1 VT case and 3 bradyarrhythmic cases. The study showed significant correlation between disease severity and arrhythmia prevalence, revealing significant increase in arrhythmia prevalence with increasing disease severity [11/116 (9.5%); 5/37 (13.5%); 8/34 (23.5%) for moderate, severe, and critical severity, respectively, p< 0.001] with a very low prevalence of arrhythmias among patients with mild COVID-19 disease (4/203; 2%). Multivariate forward regression analysis showed that background CHF and disease severity are independently associated with overall arrhythmias while age, background CHF, disease severity, and arrhythmic symptoms (syncope or palpitations) are associated with tachy-arrhythmias. Importantly, a classification tree using specific age (>70 years old) and troponin (≥ 48 ng/L) cutoffs, as well as disease severity, and history of CHF could further stratify patients into high and low risk for developing arrhythmia. To our knowledge, these relevant cutoffs were not previously described and our presented classification tree is novel in this regard.
The overall arrhythmia prevalence in our study was 7.2%, which is lower than reported in some studies1,12 and similar to others.13,14 Notably, the initial study from China reported an overall prevalence of 16.7% among 138 hospitalized COVID-19 patients in Wuhan,1 but there was a significant difference of arrhythmia incidence between non-ICU and ICU patients (6.3%, vs. 44.4%; p 0.001) and their relatively high overall prevalence was clearly influenced by the high number of ICU patients (36/138, 26%) in whom there was very high prevalence of arrhythmias. Similarly, among 115 American COVID-19 patients,12 a dramatic change in arrhythmia prevalence was noticed between non-ICU and ICU patients (0% versus 27.5%, p 0.0002), with an overall arrhythmia prevalence of 16.5% given the fact that 69/115 study patients (60%) were ICU patients.12 In contrast, an Italian study focusing on 132 stable non-ICU COVID-19 patients, reported 12/132 (9%) new arrhythmias documented during the admission period.13 Our overall arrhythmic prevalence is similar to Wuhan’s non-ICU group and to the Italian stable non-ICU patients, probably reflecting the low percentage of ICU patients (9.7%) in our cohort. Notably, the arrhythmic prevalence among our ICU patients was significantly greater than our non-ICU patients (21% versus 5.7%; p 0.003), in accordance with the previous studies. Similar findings were reported in a single center study including 700 patients (11% ICU patients), revealing an overall arrhythmia prevalence of 7.5% with marked difference of arrhythmia prevalence between ICU and non-ICU patients.14 Lastly, the low arrhythmic prevalence among the non-ICU group compared with the ICU one, suggests that need for mechanical ventilation or presence of multi-organ dysfunction necessitating ionotropic support, which characterized our ”critical” severity and most of our ICU patients, have major impact on arrhythmic prevalence as was suggested by previous studies.1,6,9,12.14
AF or atrial flutter was the dominant arrhythmia in our study, occurring for 23/28 (82%) of cases with documented arrhythmia, while only a minority of these cases had previous arrhythmias (5/23, 22%). This result is in line with multiple previous studies, revealing AF to be the most prevalent arrhythmia in COVID-19 disease.1,6,12,13,14,15 The absence of previous atrial arrhythmias in most of these patients and the fact that some of these new arrhythmia was timely correlated with respiratory deterioration, may suggest an association or perhaps even a causative relation between COVID illness and arrhythmia development. Indeed, some have suggested that the increased cytokine levels among severe COVID patients may trigger atrial arrhythmias.7,8 Nevertheless, one should not interpret the above results as a definite cause and effect relation between COVID disease and AF, as most of these patients are elderly with multiple comorbidities as HTN, diabetes and IHD and most if not all have significant respiratory symptoms, all of which are well known risk factors for AF occurrence. Thus, we like others,13,14 find it hard to confirm a necessary pathophysiologic link between AF occurrence and COVID-19 infection. Regardless of whether AF is a direct result of COVID or circumstantial to the multiple comorbidities of these patients, multiple publications suggest AF to predict bad prognosis among COVID-19 patients.8,12,14,16
There were 3 cases with new onset bradyarrhythmias in our study, including two cases with advanced AVB and one with a transient slow ventricular escape rhythm. Notably, 2 of these cases were relatively young male patients (50 and 33 year old) without prior conduction abnormalities who presented with advanced AVB or developed a slow ventricular escape rhythm, respectively. Both did not receive any negative dromotropic or chronotropic drug. Both had normal electrolytes and markedly elevated inflammatory markers with mildly increased or normal hsTnI. Accordingly, we raise the possibility, which we cannot prove, that these bradyarrhythmias had been related to the cytokine storm and exacerbated inflammatory state associated with COVID-19 illness. Interestingly, inflammatory-mediated conduction abnormalities have been suggested in previous reports of COVID-19 patients who developed otherwise unexplained AVB during the acute COVID illness, with markedly elevated inflammatory markers.8,9
Disease severity, categorized by the level of respiratory support needed, was the most robust clinical predictor for new-onset arrhythmia among COVID-19 patients in our study. This correlation was shown in most previous studies, with special attention to the remarkably high arrhythmia prevalence among COVID-19 ICU patients, many of whom were mechanically ventilated and/or with multi-organ failure necessitating inotropic support.1,2,3,12,14 The other clinical predictor for new-onset overall arrhythmias was background CHF, which is well-established risk factor for atrial and ventricular arrhythmias.17,18
Regarding laboratory predictors, previous studied suggested both myocardial injury as well as inflammatory process to have impact on arrhythmic prevalence or even underlie arrhythmia development,7,12 with the notion of ”inflammatory and cardiac injury mediated atrial tachycardia and conduction disturbances”.7,8,9 Indeed, in our study both hsTnI and CRP were significantly associated with new arrhythmias in the univariate analysis. However, in the classification tree model, hsTnI per se was found to best differentiate between low and high arrhythmic activities (Fig 5). This is in line with many previous studies suggesting myocardial injury, assessed via troponin levels, to correlate with need for mechanical ventilation, ICU admission, mortality and arrhythmias.1,2,3,4,9,19
Importantly, our classification trees (Fig 4&5) are noval and clinically relevant as they can discriminate between patients with high and low risk for new onset arrhythmia. Although, some of these variables as age and troponin are clinically intuitive and were previously reported,2,4,9,14 no cutoffs were previously given to guide clinical decisions. Accordingly, we suggest that patients whose age is below 70 with mild to moderate disease (1.6% arrhythmic prevalence) or with negative hsTnI (2.1% arrhythmic prevalence) may be hospitalized to a ”non-monitored bed” and even discharged early; while patients aged ≥70 (18.1% arrhythmic prevalence) or patients with hsTnI ≥ 48 ng/L (34.1% arrhythmic prevalence) should be carefully monitored for the occurrence of arrhythmias. These algorithms might be of crucial importance to direct optimal utilization of medical staff and monitoring equipment in COVID-19 outbreaks, especially in areas with limited medical resources.