1 INTRODUCTION

Coronavirus disease‐2019 (COVID‐19) causes viral pneumonia and severe illness resulting in organ dysfunction including acute respiratory distress syndrome (ARDS) and acute cardiac injury and renal injury. ¹ There is also an increased incidence of arrhythmias, especially in those requiring intensive care. ² As yet, few data are available regarding the effects of the infection on the electrophysiology of the heart. Electrocardiography (ECG) can be used to detect cardiac involvement in multisystem disorders, and often provides prognostic information. In COVID‐19, left bundle branch block and ST segment deviation may be associated with increased risk of mortality. ³ The effects on repolarization remain unclear. A risk of QT prolongation was highlighted in cases of COVID‐19 treated with hydroxychloroquine and Azithromycin, but the QT abnormality observed may have been more than just a pharmacological effect. ⁴

2 METHOD

We screened patients who were admitted to our institute via the emergency department with the Sars‐Cov‐2 virus infection at the height of the COVID‐19 outbreak locally. In total, 406 patients presenting to the department tested positive for the infection between 1st March – 18th May 2020. Of these, 293 patients were recruited in to the study; 106 were excluded as they did not require hospital admission and were discharged from the emergency department and seven were excluded due to the presence of a paced cardiac rhythm. Data were collected and analyzed for patient demographics, admission ECG, chest x‐ray and blood biochemical and hematological profiles. The left ventricle ejection fraction was calculated by a certified echocardiographer using the Simpson's biplane method; when this was not possible, the visual estimation was accepted. A comparison was performed of patients who succumbed to the infection (“deceased” group) and those who remained alive to the time of last accessing the hospital record at > 6 weeks after presentation (“survived” group). The study proposal was approved by the research ethics committee and complies with the principles of the Declaration of Helsinki.

Where appropriate continuous variables were conveyed as a mean ± standard deviation and categorical variables as a number and percentage. Statistical analysis was performed using the independent t‐test and where applicable, the paired t‐test to compare continuous variables. The Chi‐squared test was applied to compare categorical variables. A p‐value of < 0.05 was regarded as statistically significant. A multivariate regression analysis (Prism 9.0.0, Graphpad software, CA, USA) and a Cox‐regression analysis (SPSS v26, IBM Corp., Chicago, IL, USA) was performed for variables associated with mortality; Kaplan‐Meier curve was used for survival analysis.

2.1 ECG analysis

All patients underwent a standard 12‐lead ECG at 25 mm/s paper speed and amplitude of 10 mm/mV (MAC 5500 HD ECG system, GE healthcare, Chicago, IL, USA) on admission. This was stored in the patient electronic medical records on the hospital server. For all patients included in this study, this baseline ECG was retrieved from the electronic records and analyzed. During the study time‐frame, testing for COVID‐19 in the UK was performed only in a hospital setting, and the delay from the performance of a test to the availability of a diagnosis was 24−72 h. At the time of recording of the baseline ECG, the diagnosis of COVID‐19 had therefore not been confirmed in any of our patients, so no specific therapy had been commenced.

Basic information including rhythm, rate, QRS duration and morphology, corrected QT interval (QTc), ST segment deviation and T‐wave changes were derived from the baseline ECG. The QTc was calculated manually by a cardiologist. The QT intervals from all leads of the individual patient ECG were screened and the lead with the longest QT duration was accepted. ⁵ Measurement was from the onset of the Q‐wave to the T‐wave offset, which was the point of intersection of the downward slope tangent‐line with the baseline (Figure 1). ⁵ Correction was performed using Bazett's formula. ⁶ Corrected QT clinical limits of >450 milliseconds (ms) for males and >460 ms for females were accepted as prolonged. ⁷ ECGs showing an abnormal prolonged QTc were re‐checked by an experienced cardiac electrophysiologist. For all patients included in the study, the pre‐existing medical record was searched for a prior ECG. These pre‐COVID ECGs, when available, were analyzed in the same manner as the baseline ECG.

FIGURE 1

ECG example from the deceased group with QT measurements. a) Pre‐COVID and b) COVID admission ECG from a deceased patient. There is prolongation of the QTc with the COVID‐19 illness. The annotations demonstrate the QT interval and the R‐R interval used for correction [Color figure can be viewed at wileyonlinelibrary.com]

3 RESULTS

In this study the deceased patients (n = 133) were older (82 ± 10.5 vs 69.8 ± 17.6 years, p < 0.001) with a significantly higher proportion having a prior diagnosis of cancer (22.3% vs 13.1%, p = 0.034), ischemic heart disease (27.8% vs 10.7%, p < 0.001) and dementia (25.6% vs 10.7%, p < 0.001). They were sicker on arrival to the emergency department (4.02 vs 3.24 NEWS score, p = 0.015), were more likely to require an ITU admission (15.8% vs 6.9%, p = 0.015) and they experienced a significantly longer hospital stay (10.7 ± 8.2 vs 8.8 ± 7.9 days, p = 0.047). There was a similar proportion of male patients in both groups (55.6% vs 59.4%, deceased vs survived, respectively, p = 0.52) with similar clinical observations on admission including heart rate (91.9 vs 90.7 beats per minute, deceased vs survived, p = 0.64) and temperature (37.2°C vs 37.3°C, deceased vs survived, p = 0.728); the baseline respiratory rate was significantly faster in the deceased cohort (23.9 vs 22 breaths per minute, deceased vs survived, p  = 0.007) (Table 1).

TABLE 1

Patient demographics of both cohorts

	Deceased	Survived	p‐value
	n = 133	n = 160
Age (y)	82.0 ± 10.5	69.8 ± 17.6	<0.001
Male	74 (55.6%)	95 (59.4%)	0.52
Hospital stay (days)	10.7 ± 8.2	8.8 ± 7.9	0.047
ITU/HDU admission	21 (15.8%)	11 (6.9%)	0.015
Diabetes	34 (25.6%)	44 (27.5%)	0.71
Hypertension	84 (63.2%)	94 (58.8%)	0.061
Pre‐existing lung condition	28 (21.1%)	35 (21.9%)	0.86
Cancer	30 (22.3%)	21 (13.1%)	0.034
Atrial Fibrillation	74 (55.6%)	82 (51.3%)	0.45
IHD	37 (27.8%)	17 (10.7%)	<0.001
Dementia	34 (25.6%)	17 (10.7%)	<0.001
Stroke	21 (15.8%)	15 (9.4%)	0.096
CKD	21 (15.8%)	23 (14.4%)	0.74
LVEF (%)	52.3 ± 11.3	55.5 ± 6.4	0.053
Lowest Hb (g/L)	105.9 ± 25.1	117.7 ± 20.7	<0.001
Troponin raised	59 (44.4%)	20 (12.5%)	<0.001
Cardiac Troponin (ng/L)	1982.4 ± 5780	413.4 ± 2815	0.017
CRP (mg/L)	244.6 ± 116.4	146.5 ± 101.1	<0.001
Creatinine (μmol/L)	162.4 ± 106.4	122.9 ± 113.4	0.003
D‐Dimer (ng/mL)	5523.3 ± 5138	3872 ± 8917	0.347
On admission:
Potassium (mmol/L)	4.1 ± 0.6	4.08 ± 0.5	0.537
Magnesium (mmol/L)	0.83 ± 0.18	0.8 ± 0.12	0.314
Calcium (mmol/L)	2.32 ± 0.14	2.3 ± 0.14	0.313
Atrial fibrillation	43 (32.3%)	39 (24.4%)	0.13
Bundle branch block (>120 ms)	15 (11.3%)	14 (8.8%)	0.47
QRS duration (ms)	95.5 ± 22.9	92.1 ± 18.6	0.181
QTc (ms)	461.1 ± 37.6	449.3 ± 32.5	0.007
Prolonged QTc (>450 ms male/ >460 ms female)	61 (45.9%)	48 (30%)	0.005
Male QTc	462.1 ± 38.5	448.1 ± 33.4	0.018
Female QTc	456.5 ± 41.3	451.1 ± 31.7	0.44
Heart Rate (beats per minute)	91.9 ± 20.6	90.7 ± 21.3	0.636
Temperature (°C)	37.2 ± 0.91	37.3 ± 1.1	0.728
Respiratory rate (breaths per minute)	23.9 ± 6	22 ± 5.6	0.007
National Early Warning Score	4.02 ± 2.8	3.24 ± 2.5	0.015

The patient demographics of both cohorts in addition to the biochemistry and hematological profile during the admission. The electrolytes, ECG intervals and clinical observations were recorded at the onset of admission, on arrival to the emergency department (“On admission”).

Blood biochemistry on admission revealed a similar level of potassium (p = 0.5), adjusted calcium (p = 0.3) and magnesium (p = 0.3) in the deceased and those who survived. There was however a significantly higher level of C‐reactive protein (CRP) (p < 0.001), high‐sensitivity cardiac troponin I (p = 0.017) and creatinine (p = 0.003) noted in the deceased group. Troponin was found to be elevated above the reference range (>46 ng/l) in a significantly higher proportion of deceased patients (44.4% vs 12.4%, p  < 0.001) (Table 1).

ECG analysis of both cohorts revealed a significantly longer corrected QT interval on the baseline ECG for the patients who subsequently died (461.1 ms vs 449.3 ms, deceased vs survived, p = 0.007) in the context of a similar QRS duration (95.5 ms vs 92.1 ms, deceased vs survived, p = 0.181) and similar left ventricular function (52.3% vs 55.5%, deceased vs survived, p  = 0.053) (Table 1). For the deceased patients in whom a pre‐COVID ECG was available (n = 63), the QT interval was found to have increased significantly from the pre‐COVID ECG to the baseline ECG on presentation with COVID (448.4 vs 472.9 ms, respectively, p < 0.001); this increase was also observed in those who survived (n = 109; 435.1 vs 451 ms, pre‐COVID vs COVID, p  < 0.001) but the degree of prolongation was greater in the patients who subsequently died (Figure 2). Of the cases found to have a prolonged QTc, troponin was higher in the deceased patients compared to those who survived, although this difference did not reach statistical significance (2002.5 vs 1115.7 ng/L, respectively, p = 0.52).

FIGURE 2

Individual changes of the QTc. (A) There is an overall trend of prolonging QTc with the Covid‐19 infection. There was significant prolongation of the QTc from pre‐COVID measurement in the (B) deceased (p < 0.001) and the (C) survived patients (p < 0.001). The degree of prolongation was notably greater in the deceased patients [Color figure can be viewed at wileyonlinelibrary.com]

3.1 Predictors of mortality

Multivariate regression analysis was performed for predictors of mortality. This revealed that age (OR 1.11 [1.07‐1.16], p < 0.001), CRP (OR 1.013 [1.008‐1.02], p < 0.001) and QTc (OR 1.02 [1.0096‐1.04], p  = 0.001) were significantly associated with an increased risk of mortality. A Cox‐regression analysis was also performed which identified Age and CRP as independent variables associated with mortality. Prolonged QTc was also highlighted as an independent predictor in this analysis (Table 2). All patients with a prolonged QTc in the fourth quartile (>469 ms) were associated with an increased risk of mortality comparatively to patients in the lower quartiles (p  = 0.006, HR 1.98 [1.2‐3.2]) (Figure 3). Gender discrimination revealed that patients exhibiting a corrected QT‐prolongation exceeding dichotomy limits of >455 ms (male) and >465 ms (female) were less likely to survive compared to patients with a normal QTc (p  < 0.01, HR 1.78 [1.2‐2.6]) (Figure 4).

TABLE 2

Multivariate regression analyses for mortality

Multivariate regression analysis
Variable	Odds Ratio	UCL	LCL	p‐value
Age on admission	1.113	1.16	1.073	<0.0001
C‐reactive protein	1.013	1.019	1.008	<0.0001
ECG: prolonged QTc	1.022	1.04	1.0096	0.001
Multivariate Cox‐regression analysis
Variable	Hazard ratio	UCL	LCL	p‐value
Age on admission	1.04	1.062	1.027	<0.001
C‐reactive protein	1.003	1.001	1.004	0.002
*ECG: prolonged QTc	1.49	2.127	1.04	0.028

*prolonged QTc: >455 ms in male, >465 ms in female patients.

Prolonged QTc was a significantly independent variate associated with mortality, reducing the odds of survival. In the Cox‐regression analysis, prolonged QTc (>455 ms [male] and >465 ms [female]), age and C‐reactive protein were also independently associated with mortality.

FIGURE 3

Cox‐survival plot comparing mortality risk in QTc quartiles. In our cohort, a prolonged corrected QT‐interval in the fourth quartile (>469 ms) was associated with an increased risk of mortality when compared to the lower quartiles (p = 0.006, HR 1.98 [1.2‐3.2]) [Color figure can be viewed at wileyonlinelibrary.com]

FIGURE 4

Kaplan‐Meier survival curve of patients with a prolonged QTc and normal QTc. At 6 weeks from the date of admission, the probability of survival in patients with a prolonged QTc (>455 ms male; >465 ms female) was significantly lower than those patients who had a normal QTc (p = 0.0027, HR 1.78 [1.2‐2.6]) [Color figure can be viewed at wileyonlinelibrary.com]

4 DISCUSSION

We have shown that the QTc interval lengthens during COVID‐19 infection, and that the degree of lengthening is greater in those who subsequently die during the illness. This finding was based on the admitting ECG before exposure to any pharmaceutical known to prolong the QT‐interval, and is suggestive of mechanisms beyond cardiotoxicity that may alter ventricular repolarisation. ⁴ , ⁸ Complementing previous reports of ECG abnormalities in COVID‐19, ³ , ⁴ this ECG anomaly may also be associated with poor outcome and therefore the findings of this study could have meaningful clinical implications.

The mechanism by which the COVID‐19 infection prolongs ventricular repolarization remains speculative: The Sars‐Cov‐2 virus has a high affinity for the ACE2 receptor and through its occupancy of this receptor can lead to over‐expression of angiotensin‐II. ⁹ An inhibitory effect exerted by angiotensin‐II on the rapidly activating components of myocyte delayed rectifier potassium current (I_Kr ) has been described previously and may result in prolonged action potential duration, thus QT prolongation. ¹⁰ Cardiac injury may also be a contributing factor as it is an established feature of COVID‐19 infection ⁸ and is associated with an increased incidence of ventricular arrhythmia. ¹¹ In our study cardiac injury was more prevalent in the deceased cohort, as evidenced by a higher degree of troponin leak in this group (1982.4 vs 413.4 ng/L, p = 0.017) and greater proportion of cases with a troponin level raised above the reference range than the patients who survived.

COVID‐19 associated QT‐prolongation may also be a secondary feature, reflecting the severity of a multi‐system illness. In our study, the deceased patients were significantly more unwell on‐admission as indicated by their higher National Early Warning Score (NEWS), a United Kingdom adopted scoring system implemented to detect patients requiring critical care. ¹² The intense inflammatory response characteristic of COVID‐19 could produce QT prolongation both by direct and indirect mechanisms. ¹¹ The up‐regulation of angiotensin‐II by Sars‐Cov‐2 via binding to the ACE2 receptor can result in excessive inflammation and injury ⁹ ; the QTc could be an expression of the degree of inflammation. This is consistent with reports which have indicated that ITU stay was associated with QT prolongation ¹³ and patients admitted to ITU were disposed to arrhythmias. ¹ , ²

The multivariate analysis revealed that prolonged QTc on presentation was independently associated with an increased risk of mortality in COVID‐19 (p  = 0.001). This finding has rarely been reported and could easily have been overlooked. ³ , ¹⁴ Romero et al demonstrated that new T‐wave inversion in COVID‐19 was associated with mortality. ¹⁴ The ECGs reproduced in their manuscript to illustrate this T‐wave inversion also demonstrate marked prolongation of the QTc, although this finding was not remarked upon and the interval was apparently not measured. Lanza et al did find a prolonged QT in a small proportion (5.8%) of COVID‐19 patients. ³ By treating QT prolongation as a binary variable, we believe that previous authors may have missed instances in which the interval was prolonged beyond the level that was normal for that patient, but did not exceed the range defined as normal for the population. By contrast, we measured the QTc in all patients and compared the results both to the reference range and when available, to the pre‐COVID value in the same patient. This difference in methodology may have given us a greater capacity to detect the effects of COVID‐19 on ventricular repolarization.

QT‐prolongation is not a binary event and the accepted definitions of a prolonged QTc (>450 ms in male; >460 ms in female) ⁷ are designed for application in a healthy population for predicting risk of ventricular arrhythmia. In our cohort, patients expressing a QTc in the fourth quartile (>469 ms) were likely to succumb to the illness. This is in keeping with prior evidence. Previous large observational studies showed a significantly increased all‐cause mortality risk at >460 ms, ¹⁵ , ¹⁶ with accentuated risk as the QT lengthens further. ¹⁷ , ¹⁸ Using the Cox‐regression analysis, we also identified QTc limits of >455 ms in males and >465 ms in females as the gender discriminated optimum dichotomy points for the association with mortality. This gender difference of 10 ms is also consistent with the literature ⁷ and studies with gender QTc specificities have indicated an augmented mortality risk in males with a QTc of >450 ms ¹⁸ and in females with a QTc ranging over 460 ms ¹⁹ and 470 ms. ¹⁸

The QT‐interval responds to changes in heart rate ²⁰ and body temperature ²¹ and is sensitive to electrolyte disturbance, owing to its dynamic nature. All methods of QT interval correction for heart rate are vulnerable to inter‐individual variation and all have proved inaccurate at extremes of the heart rate spectrum. Within our series the groups are well matched for the baseline heart rate and metabolic state, and therefore it is unlikely these factors can explain the differences in the QT interval. The association between QTc prolongation and mortality is also unlikely to be affected by the heart‐rate correction formula used. ¹⁶

5 CONCLUSION

Prolongation of the QTc interval is associated with mortality in COVID‐19 infection and appears to be independent of drug‐toxicity. Frequent QTc monitoring on ECG is advocated in the risk stratification of COVID‐19 patients.

CONFLICTS OF INTEREST

Mark Gallagher has received research funding from Attune Medical and has acted as a consultant and a paid speaker for Boston Scientific and Cook Medical.

Sumeet Sharma has received educational grants and has acted as a paid speaker for Boston Scientific, BAYER, Abbott medical, Pfizer and Bristol‐Myers Squibb.

AUTHOR CONTRIBUTIONS

All the authors have contributed significantly to this work. Data collection was completed by Zaki Akhtar, Victoria Ewasiuk and Louise Gregory. The corrected QT measurements were verified by Dr Mark M. Gallagher and the methodology by Dr Yee Guan Yap and Dr Zhong Chen. Data analysis was performed by Zaki Akhtar, Ahmed I. Elbatran and Sumeet Sharma. The manuscript drafting was undertaken by Zaki Akhtar, Mark M. Gallagher, Yee Guan Yap, Lisa WM Leung, and Sumeet Sharma. Critical review was performed by all the authors, in particular Brendan Madden, Aodhan Breathnach, Zhong Chen and David S. Fluck. Final approval received significant contribution from all the authors.

FUNDING

Zaki Akhtar has received funding for a research fellowship from Abbott Medical.