Predictive Value of Electrocardiogram for the Occurrence of Major Adverse Cardiac Events in Patients with Pulmonary Embolism

Pulmonary embolism (PE) is one of the most prevalent cardiovascular diseases worldwide. A few studies have advocated the applicability of electrocardiogram (ECG) for the determination of prognosis of PE patients. Considering the low-cost and wide availability of ECG we aimed to investigate the association of selected ECG parameters with the occurrence of major adverse cardiac events (MACE) in PE patients. In this study, 733 adult patients admitted with a denite diagnosis of acute PE were included from a registry of PE patients in a tertiary heart center. The patients’ clinical records were retrospectively reviewed, and demographic information, ECG abnormalities as well as the on (including mechanical Syncope, cardiogenic shock, or in-hospital were


Introduction
Pulmonary embolism (PE) is one of the most prevalent cardiovascular diseases with an incidence of 39-115 per 10,000 population (1)(2)(3). Thrombolytic drugs are administered in patients with PE presenting with cardiogenic shock. However, low molecular weight or unfractionated heparin are considered as the therapeutic choices for hemodynamically stable patients (1). Therefore, an optimal selection of therapies for PE patients relies on risk strati cation and identifying those in need of more aggressive types of treatment to improve prognosis. Right ventricular dysfunction determined by transthoracic echocardiography or elevated biomarkers (cardiac troponin or natriuretic peptide) have been suggested by some studies to have a signi cant association with prognosis in PE patients (4)(5)(6). However, transthoracic echocardiography is not widely available in all centers and interpretation is largely operator-dependent.
About 70 percent of patients with PE have abnormal electrocardiograms (ECG)s, which comprise a wide range of changes including arrhythmia (supraventricular or ventricular tachycardia), conduction abnormalities (right bundle branch block [RBBB]), deviation of the electrical axis of the heart (left or right axis deviation [LAD and RAD]), changes in the P-wave shape, QRS wave amplitude, and QT dispersion (7).
Previous studies have investigated various ECG parameters in the diagnosis of PE, but the results were not consistent (8)(9)(10). Therefore, ECG ndings are not currently considered speci c for the diagnosis of PE. However, some studies have advocated the applicability of ECG for the determination of PE prognosis. In a study carried out by Akgüllü et al., QT interval dispersion and P wave dispersion were signi cantly associated with early death in acute PE (11). Moreover, Escobar et al. proposed that sinus tachycardia and atrial arrhythmia were independent predictors of a poor prognosis in hemodynamically stable PE patients (12). Considering the low cost and wide availability of ECG, we aimed to investigate some parameters of ECG that have been previously underappreciated in PE patients and to examine their association with the occurrence of major adverse cardiac events (MACE).

Methods
In this study, 1,064 patients (age range: 18-80 years) admitted with a de nite diagnosis of acute PE between April 2011 and April 2017 were included from a registry of PE patients in a tertiary heart center (center name hided for peer review). The diagnosis of patient must have been con rmed by computed tomography (CT) scan or lung perfusion scan to be included in our study. The exclusion criteria were antiarrhythmic medication use such as digoxin, severe metabolic diseases such as hypokalemia, myocardial infarction, heart failure, angina pectoris, sepsis, congenital heart disease, cor pulmonale, presence of left bundle branch block (LBBB) in admission ECG, pacemaker, and lack of an interpretable standard ECG in the rst 24 hours of the onset of symptoms.
The study was conducted in accordance with the declaration of Helsinki. The protocol of the study was approved by the medical ethics committee of the university (university name hided for peer review).
Written informed consent was obtained from all patients for entering their data into the registry of patients with pulmonary embolism.

Measurements and de nitions
The patients' clinical records were retrospectively reviewed, and demographic information, including age and gender, were recorded in the prepared checklists. Data on MACE was extracted from patients' les.
MACE was de ned as the occurrence of at least one of the following events: Hypotension: Systolic blood pressure (BP) less than 90 mmHg Mechanical ventilation Syncope: A transient loss of consciousness due to insu cient blood ow to the brain Cardiogenic shock: A signi cant reduction in BP (systolic BP < 90 mmHg) lasting more than 30 minutes and accompanied by symptoms of end-organ hypoperfusion In-hospital mortality due to PE complications.

ECG analysis
We analyzed the patients' ECG, which was obtained in the rst 24 hours since the onset of symptoms using a supine, standard 12-lead ECG at 25 mm/s paper speed, and 10 mm/mV amplitude by MAC 500 ECG machine (GE medical system, USA). The investigations of ECG parameters were conducted manually with the help of a magnifying glass by two experienced cardiologists who were blind to the clinical data of the patients. Three consecutive beats were investigated for the analyses, where at least 10 leads were analyzable.
The following ECG changes were recorded in the checklist: ST-T changes in precordial leads, S1Q3T3, RBBB, incomplete RBBB (iRBBB), ST elevation in lead V1, ST elevation in lead III, ST elevation in lead aVR, low voltage QRS, QR wave in lead V1, and inverted T wave in leads V1 to V4.
Inverted T wave was de ned as a T wave > 2 mm below the isoelectric line in two or more adjacent leads (V1 to V4).

Statistical analysis
The analysis was performed by Stata version 16 (StataCorp, USA). The Kolmogorov-Smirnov test was used to analyze the normality of quantitative variables. For those variables with normal distribution, we made comparisons between two groups by independent-sample t-test, and the descriptive statistics were presented as mean ± SD. For numeric variables that were not normally distributed, we made comparisons between two groups by the Mann-Whitney U test, and the descriptive statistics were presented as median (25-75 percentiles). To analyze categorical data, we used a chi-square test and reported the Kappa value, and the descriptive statistics were presented as frequency (percentages). Multiple logistic regression analysis was performed to investigate the association between study variables and the occurrence of the MACE.
Furthermore, we used the receiver operating characteristic (ROC) curve to determine the area under the curve (AUC). Accordingly, the sensitivity, speci city, likelihood ratio (+), likelihood ratio (-), positive predictive value, and negative predictive value were reported. The P values less than 0.05 were considered as statistically signi cant.

Results
Of 1,064 patients with PE included in the study 331 patients were excluded (aged over 80 years in 107 patients, lack of CT angiography or perfusion scan in 110 patients, low EF in 41 patients, LBBB in 28 patients, non-interpretable ECG in 14 patients, acute on chronic PE in 14 patients, permanent pace maker in 5 patients, digoxin use in 4 patients, cor pulmonale in 2 patients, congenital heart disease in 2 patients, other critical diseases in 2 patients, and recent myocardial infarction (MI) and septic embolism in 2 patients). The nal population consisted of 733 PE patients with a mean age of 57.60 ± 15.73 years. Of these, 379 patients (51.7%) were male, and 354 patients (48.3%) were female.
The investigated ECG parameters in PE patients are described in Table 1. The most common sign was the presence of an inverted T wave, which was seen in 306 patients (41.9%). Other common signs were S1Q3T3 in 283 patients (38.6%), ST elevation in aVR in 206 patients (28.2%), iRBBB in 138 patients (18.9%), and ST elevation in V1 in 123 patients (16.8%). In majority of patients (91.4%), the ratio of S wave to R wave in lead I was more than 1. Logistic regression analysis showed that age (P = 0.395) and sex (P = 0.150) had no signi cant relationship with the occurrence of MACE.
The results of multivariate regression analysis on the relationship between the ECG signs and the occurrence of MACE are described in Table 2. ST-elevation in aVR (OR = 3.87, 95%CI = 2.32-6.44, P = 0.001) and S1Q3T3 (OR = 2.04, 95%CI = 1.22-3.43, P = 0.007) were independent predictors of MACE. Other ECG signs in patients with PE were not signi cantly associated with the occurrence of MACE. The ROC curve was plotted for two variables that had a signi cant relationship with the occurrence of MACE (Fig. 1). The results of ROC analysis and the predictive values of ST elevation in the aVR lead and S1Q3T3 for the occurrence of MACE are described in Table 3. The predictive values of ST elevation in the aVR lead and S1Q3T3 for the occurrence of MACE were moderate (the AUC for ST elevation in the aVR lead = 0.664, and for S1Q3T3 = 0.640). The sensitivity of ST elevation in the aVR lead and S1Q3T3 was 53.1 and 59.9%, and speci city was 79.7 and 68.1%, respectively.

Discussion
According to the guidelines of the European Society of Cardiology (ESC), the risk strati cation and determination of the prognosis of the PE patients is necessary and can be helpful in choosing the best therapeutic strategy (1). However, the applicability of ECG changes for determining the prognosis of PE patients is under debate. Inverted T wave in precordial leads, S1Q3T3, and ST elevation in aVR were the most common ECG parameters in PE patients. These ndings were also previously remarked as the most common ECG abnormalities in PE patients (9,(13)(14)(15). Furthermore, the results of the current study showed that some changes in ECG, including RBBB, iRBBB, S1Q3T3 sign, ST elevation in leads V1 and III, QR wave in lead V1, and inverted T wave in precordial leads were signi cantly more frequent in the MACE group than the control group. ST elevation in lead aVR and S1Q3T3 sign were associated with a higher risk of MACE during hospitalization. Although the sensitivity and positive predictive value of these changes were low, the speci city and negative predictive value were high. Therefore, these parameters in ECG can be considered as a cheap and widely available tool for rolling out the PE patients who are not at high risk of MACE. A similar study by Kukla et al. on 292 acute PE patients also reported that atrial brillation, S1Q3T3 sign, negative T waves in leads V2-V4, ST-segment depression in leads V4-V6, STsegment elevation in leads III, V1 and aVR, QR in lead V1, RBBB, higher number of leads with negative T waves, and higher sum of the amplitude of negative T waves were more frequent in patients who experienced PE complications (7). Moreover, in multivariate analysis, ST-segment elevation in leads aVR (OR 2.49; p = 0.011) was identi ed as an independent predictor of complications during hospitalization. Nevertheless, this study did not report the sensitivity and speci city of these parameters for predicting the occurrence of complications during hospitalization (7). Also, Janata et al. investigated 396 PE patients and revealed that ST elevation in aVR was the only signi cant predictor of mortality in the intermediaterisk group. However, this association was not detected in high-risk patients (16). In a larger study conducted on 508 PE patients, Geibel et al. demonstrated that the presence of at least one of the following ECG ndings, besides hemodynamic instability, syncope, and pre-existing chronic pulmonary disease, was a signi cant independent predictor of early (30-day) mortality. These ECG ndings were atrial arrhythmias, complete RBBB, peripheral low voltage, Q waves in leads III and aVF, and ST-segment elevation or depression over the left precordial leads. In this study, the association of ECG ndings with inhospital mortality was not evaluated individually (17). Moreover, some prior studies used a combination of different ECG signs to develop a scoring system. Toosi et al. investigated a 21-point system based on ECG changes in 159 acute PE patients and reported that having a score of ≥ 3 in this scale can predict the occurrence of right ventricular dysfunction with high sensitivity and speci city of 76 and 82 percent, complicated disease course with moderate sensitivity and speci city of 58 and 60 percent, and mortality incidence with moderate sensitivity and speci city of 59 and 58 percent, respectively (10). Iles et al. evaluated a similar 21-point scoring system based on ECG changes on 229 PE patients and reported that an ECG score of ≥ 3 predicted those with > 50% perfusion defect with a sensitivity of 70% (95% con dence interval [CI], 59 to 81%), and a speci city of 59% (95% CI, 51 to 67%) (18). Another study also used a 21-point ECG score to predict right ventricular dysfunction, and reported a sensitivity of 92% and a negative predictive value of 97%; similarly, complications during hospitalization were predicted with sensitivity and negative predictive value of 75% and 92%, respectively (19). These studies suggested that ECG can be considered as s useful tool for the determination of the prognosis of PE patients; however, none of them investigated the association of the ECG ndings individually with PE prognosis.
As far as we investigated, our study was the largest study that evaluated the prognosis of PE patients based on their ECG ndings. Moreover, we reported sensitivity, speci city, and predictive value as well as OR and 95% CI for each ECG abnormalities individually in order to better clarify the association of ECG ndings with the occurrence of MACE to be used in the clinical setting. However, our study had some limitations. We did not evaluate the laboratory data, medical and drug history, clinical presentation, echocardiographic ndings, and medications administered during hospitalization for all the patients. This was partially due to some missing data of the patients' les. As a result, we did not adjust the outcome for these confounding variables, which can be considered as a possible source of bias for our ndings.
Moreover, we excluded the patients who had a positive history of taking antiarrhythmic medication such as digoxin, severe metabolic disease such as hypokalemia, myocardial infarction, heart failure, angina pectoris, sepsis, congenital heart disease, cor pulmonel, LBBB in admission ECG, and permanent pacemaker. These patients were excluded due to the possibility of ECG changes caused by these abnormalities that could interfere with the ECG changes due to PE. However, the exclusion of these patients can partially limit the generalizability of our ndings as a considerable proportion of the PE patients have at least one of these abnormalities.

Conclusion
Some parameters of ECG, such as S1Q3T3 and ST elevation in aVR, are independent predictors of MACE in patients with PE. Investigation of these parameters in ECG can be considered as a cheap and widely available strategy for the determination of the prognosis of PE patients, in particular for rolling out those who have less possibility of experiencing MACE.

Declarations
Ethics approval and consent to participate: The study was conducted in accordance with the ethical guidelines of the declaration of Helsinki and the protocol was approved by the ethics committee of the university.