Acute pulmonary thromboembolism (APTE) is the third most frequent cause of cardiovascular morbidity and mortality after acute coronary syndrome and stroke. With improvements in diagnostic modalities and physician awareness, the inpatient mortality of APTE in China decreased progressively from 25.1% in 1997 to 8.7% in 2008 [1]. The clinical presentations of APTE varies, ranging from asymptomatic and incidentally discovered emboli, to massive embolism causing sudden death prior to diagnosis. In Europe, 34% of pulmonary thromboembolism (PTE) patients suffer immediate death, 59% remain undiagnosed until death, and only 7% having a definitive diagnosis before death [2]. Hence, timely diagnosis of PTE and screening out massive PTE can be challenging, and the overall mortality from massive pulmonary embolism (PE) remains as high as 52% [3].

Electrocardiographic (ECG) changes are increasingly reported as being useful for assessing the severity of PE, while also showing a strong correlation with hemodynamic collapse of right ventricular dysfunction [4, 5, 6, 7, 8, 9, 10]. So far, there is a notable absence of ECG in the clinical tools available for prognostication of PE in the guidelines. To address this, Daniel et al. [11] reported an ECG scoring system in 2001 that could predict the severity of pulmonary artery hypertension (PAH) in PTE and the degree of perfusion defect on ventilation-perfusion lung scanning. However, few Chinese studies have so far evaluated the utility of ECG abnormalities for assessing clot burden, hemodynamic status, right ventricular function, survival rates and management of patients presenting with acute PTE. The initial aim of this study was therefore to develop a revised scoring system that could predict the severity of acute PTE (localization and quantification) on the basis of ECG abnormalities. This should assist clinicians when screening patients who present with massive PTE prior to imaging diagnosis.

In the present study we found that the presence of the S1Q3T3 pattern, sinus tachycardia and RAD were associated with significantly heavier thrombotic load and were more frequent in patients with pulmonary trunk embolism than peripheral embolism. Furthermore, multivariate regression analysis revealed that sinus tachycardia was an independent predictor of thrombotic load (odds ratio (OR): 2.42; p = 0.016). In addition, we observed that RBBB, atrial arrhythmias and V1–V3 TWI showed trends for correlation with the Qanadli score or with the location of thrombus. There is mounting evidence to suggest these three ECG abnormalities are more frequent in cases with massive PTE than cases with sub-massive or no PTE, while still being valuable in the prognostic assessment of PTE [5, 6, 13, 14, 15, 16, 17, 18, 19, 20]. These paradoxical results may be due to our failure to screen out the new-onset RBBB, atrial arrhythmias and TWI in V1–V3, or because of the small sample size.

ECG changes in PTE essentially relates to dilatation of RV and alteration in the contractile properties of the RV myocardium induced by abrupt increase in pulmonary vascular resistance (PVR) [21]. Other pathophysiologic mechanisms that contribute to the ECG changes include PTE-induced release of inflammatory mediators, RV strain and hypoxia [22]. After reviewing 10 studies comprising 3007 patients with APTE, Shopp et al. [23] concluded that six findings of RV strain on 12-lead ECG (tachycardia, S1Q3T3, complete RBBB, TWI in V1–V4, ST elevation in aVR, and AF) were associated with an increased risk of circulatory shock and death.

Indeed, tachycardia is the only ECG abnormality included in both the original and simplified versions of the pulmonary embolism severity index (PESI) score [24, 25]. According to Kucher et al. [26], a heart rate $>$100 bpm without specifying the presence of sinus tachycardia or supraventricular tachycardia is correlated with an escalation of therapy [6]. Kukla et al. [20] showed that atrial fibrillation was present in 231 of 975 (24%) patients with APTE, and was associated with a higher risk of mortality (23% vs 12%, OR 2.1, p 0.001) and complications (31% vs 20%, OR 1.8, p 0.001). The S1Q3T3 pattern is one of the most typical ECG findings and is found more frequently in the setting of acute massive PTE [27]. Several studies have also found that RAD on ECG was associated with the severity of PTE and with a higher risk of adverse in-hospital courses [28, 29, 30], although others found no significant difference in this regard [6, 9, 14]. In addition, the frequency of RBBB in association with APTE has been reported to range from 6% to 69% [16, 31, 32, 33, 34, 35]. In the present study, RBBB was found in 12.2% of the derivation group and 10.5% of the validation group. Of note, PTE-related TWI is a repolarization abnormality that has been consistently reported as the most common ECG abnormality, with a variable frequency from 16% to 82.9% [17, 33, 36, 37, 38]. The PTE-related TWI frequency in the present study was 12.2%, which is lower than the frequencies reported in western countries. Although a uniform consensus has yet to be established for association of the above ECG abnormalities with adverse prognosis (including elevated cardiac biomarkers, RV enlargement, in-hospital complications or mortality), the potential utility of ECG findings as prognostic indicators for the severity of PTE is now widely recognized.

Rodrigues B, et al. [39] showed that a Qanadli score $>$18% was correlated with RV dysfunction on echocardiography (OR 10.85; 95% CI 3.20–36.7; p 0.001). It is in line with the cut-off value obtained in the present study (25%), and lower than that obtained by Qanadli et al. [12] (40%) and Wu et al. [40] (60%). The result reflects differences in the populations studied.

To help distinguish patients with massive PTE from those with smaller PTE, we developed the CHARIS I and CHARIS II algorithms to calculate the risk of a Qanadli score $>$25% and the probability of trunk obstruction, respectively, in individual patients. Our models firstly combine the ECG findings with the comorbid diseases of COPD, HTN, CHD, CHF and AF. The risk prediction models undergo both internal and external validation processes. As expected, both algorithms were well suited for identifying patients in the derivation cohort with a heavy thrombus burden, as shown by a C-statistic of 0.72 and 0.75 for CHARIS I and CHARIS II, respectively. However, the prediction models showed slightly less discrimination in the external validation cohort. This could be interpreted by the imbalance of baseline characteristics between the two cohorts, based on the fact that patients in the derivation group had a lower mean Qanadli score (27.5% vs 42.0%, p 0.001), while a greater proportion of patients in the validation group had trunk obstruction (42.1% vs 23.2%, p = 0.002). On the other hand, the calibration plot and Hosmer-Lemeshow test further confirmed and reinforced the accuracy of the prediction models in the external validation sample. Moreover, the DCA results suggest the rules were clinically useful, making them effective and feasible tools for the early identification of patients with a high Qanadli score or trunk obstruction prior to lung perfusion scintigraphy or pulmonary arteriography.

Currently, MCTPA is the preferred diagnostic modality for outpatients and emergency patients with suspected PTE. However, MCTPA is contraindicated for patients with history of severe contrast allergy, severe renal insufficiency, hyperthyroidism, or who are pregnant. Moreover, it is not available in remote areas and primary hospitals. 12-channel ECG is the most affordable, simple, secure, rapidly interpretable, repeatable and noninvasive test for the diagnosis and risk assessment of PTE. In summary, the present study used ECG signs and removed the impacts of several comorbidities on ECGs to develop two novel scoring systems that can help to estimate the clot burden of PTE and decisions regarding hospitalization.

Several limitations of our study should be addressed. Firstly, despite being a multicenter study, the sample size of the derivation and validation cohorts was relatively small, and validation with a larger patient cohort is needed. Secondly, other ECG findings that were not included in our models, such as ST-segment depression (STD) [4, 6, 8, 10, 14], ST Elevation [6, 8, 10, 14, 41, 42], low QRS voltage in the peripheral leads [4, 8], QRS Fragmentation [8] and long QT [43]. These ECG changes were reported to provide valuable prognostic information in acute PTE. Therefore, the use of our rules to assist with diagnosis may result in some patients with massive PE being missed. Additionally, the severity of PTE in our study was defined by the degree of obstruction or filling defect on MCTPA. However, this is not in accordance with the current definition of massive PTE, which is characterized by hemodynamic collapse. Further research is needed to evaluate the accuracy of our algorithms for predicting the risk of adverse outcomes in patients with APTE. Another weakness of the study was that our models were not compared to previous scoring systems such as the Daniel score. Finally, patients with incomplete data were excluded on account of retrospective design, which may have introduced some patient selection bias. Despite these limitations, the greatest strength of the present investigation was the incorporation of several comorbid diseases into the models, thus reducing the confounding bias introduced by these comorbidities on the ECGs.

Algorithms were developed based on ECG abnormalities for quantitative and localization evaluation of thrombus burden in patients with APTE and then validated in an external population. The CHARIS I and CHARIS II models showed moderate discrimination and good calibration, while also demonstrating positive net benefits by decision curve analysis. Taken together, the novel algorithms were shown to be useful tools that could assist physicians when screening patients with massive PTE prior to imaging diagnosis, as well as in making decisions regarding patient hospitalization.

The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.

FW, LW, DCX and YWX designed the research study. FW, CXY, XXC, HPW and KYZ made contributions to the data acquisition. FW and DCX analyzed the data. All authors contributed to editorial changes in the manuscript. All authors read and approved the final manuscript. All authors have participated sufficiently in the work and agreed to be accountable for all aspects of the work.

All patients provided written informed consents, the study was approved by the competent ethics committee (SHSY-IEC-4.1/21-179/01) and complied with the Declaration of Helsinki and good clinical practice guidelines.

Not applicable.

This study was funded by Clinical Research Plan of Shanghai Hospital Development Center (No. SHDC2020CR3030B).

The authors declare no conflict of interest.