We retrospectively enrolled 241 patients with laboratory-confirmed COVID-19 who were admitted to Renmin Hospital of Wuhan University from January 11 to March 2, 2020. COVID-19 was diagnosed with a positive result by real-time reverse transcriptase–polymerase chain reaction (RT-PCR) from clinical samples. Exclusion criteria were patients whose CT images were difficult to analyze because of poor quality. The study was approved by the Research Ethics Committee of the Renmin Hospital of Wuhan University, Wuhan, China (approval no. WDRY2020-K038). Written informed consent was waived by the ethics commission as it is a retrospective study. The procedures are in accordance with the Helsinki Declaration and the ethical standards of the human experiment responsibility committee.

The sample size was estimated based on the primary difference between the diameter of the left atrium (LAD) + left ventricle (LV) measurements. We assumed the mean and SD of the major adverse cardiovascular event (MACE) group were 12.1 and 1.5, respectively. The mean and SD of the control group were 11.4 and 1.3, respectively. The probability of type I error was 0.05 (with both sides), and the probability of type II error was 0.20. We calculated the necessary sample size for this research to be 126.

Patient data including demographics, clinical characteristics, laboratory examinations, imaging findings, comorbidities, complications, treatment measures, and outcomes were collected from electronic medical records by two investigators (Xiaoyang Zhou and Tingting Liu). Chest CT was done for all the patients and reviewed by 2 radiologists independently (Shuang Li and Minjie Lu). The severity of COVID-19 was defined according to the Chinese management guideline for COVID-19 (version 7) (9). A patient was classified as “mild” if he had mild symptoms, with or without pneumonia on imaging. Patients who met any of the following criteria will be classified as “severe”: (1) shortness of breath, respiratory rate ≥30 times/min; (2) Sao₂ ≤93% (at rest); (3) Pao₂/Fio₂ ≤300 mm Hg (1 mm Hg = 0.133 kPa). Patients who met any of the following criteria will be classified as “critical”: (1) respiratory failure that needs mechanical ventilation; (2) shock; (3) combined with other organ failure. Clinical outcomes were monitored until April 10, 2020. The endpoint was COVID-19–associated death. The discharge criteria included normal body temperature for more than 3 days, obvious improvement in respiratory symptoms and chest imaging, and at least two negative results for RT-PCR for COVID-19 (at least 24 h between sampling) (9). According to whether adverse cardiovascular events occurred during the follow-up, they were divided into two groups. The MACEs were heart failure, ischemic cardiovascular events, and malignant arrhythmia. Ischemic cardiovascular events include acute myocardial infarction and recurrent angina. Malignant arrhythmia was defined as rapid ventricular tachycardia lasting more than 30 s inducing hemodynamic instability and/or ventricular fibrillation. Cardiac injury was defined as blood levels of cardiac biomarkers [high-sensitivity TnI (hs-TnI)] above the 99th-percentile upper reference limit (0.04 ng/mL), regardless of new abnormalities in electrocardiography and echocardiography (10).

Non-contrast chest CT scan was obtained with the patient in the supine position, and scanning was performed at end inspiration. To minimize motion artifacts, patients were instructed on breath-holding. CT examinations were done using Optima CT 680 (GE Healthcare, Bright Speed). The scans were acquired and reconstructed as axial images using the following parameters: 5.0-mm section thickness, 5.0-mm interval, 120 kVp, and adaptive tube current.

The CT images were assessed for the following features or patterns: (1) ground-glass opacity (GGO): hazy areas with slightly increased density in lungs without obscuration of bronchial and vascular margins (11); (2) consolidation: an increase in pulmonary parenchymal density with obscuration of bronchial and vascular margins, because alveolar air is replaced by pathological fluids, cells, or tissues (11); (3) reticular pattern: a collection of innumerable small linear opacities on CT images due to thickened pulmonary interstitial structures such as interlobular septa and intralobular lines (11, 12); (4) crazy paving pattern: thickened pulmonary interstitial structures with superimposition on a GGO background, resembling irregular paving stones (12); (5) air bronchogram: a pattern of air-filled bronchi on a background of opaque airless lung (11); (6) airway changes: include bronchiectasis and bronchial wall thickening; (7) pleural changes: including pleural thickening and pleural effusion; (8) fibrosis; (9) air bubble sign: a small air-containing space in the lung (12); (10) nodules: a rounded or irregular opacity with well- or poorly defined edges, measuring less than 3 cm in diameter (11); (11) halo sign: nodules or masses surrounded by ground glass (11); (12) lymphadenopathy; (13) pericardial effusion (Figure 1).

The anteroposterior LAD and the length of the LV were measured. The sum of the two values was used as an index to evaluate the size of the heart. The cardiothoracic ratio (CTR) was also calculated. The slice in which the heart is the largest was selected; the maximum transverse diameter of the heart and the largest transverse diameter of the chest were measured, and the ratio of the two was calculated (Figure 2). We also measured the CT attenuation value of the myocardium: the median axial slice first (level 2) was select, and then the slices 4–10 mm above the median axial slice (level 1) and 4–10 mm below the median axial slice (level 3) were selected. Three regions of interest (ROIs) were drawn in these slices: the middle of the interventricular septum, the anterior wall of LV and the lateral wall of LV. The ROI should be round, with an area of 0.05 cm². The mean, maximum, and minimum values of the myocardium were obtained (Figure 2).

Continuous variables were expressed as mean [standard deviation (SD)] or median [interquartile range (IQR)], and categorical variables were expressed as proportions. Student t-test, Mann–Whitney U test, and χ² test or Fisher exact test were used to compare differences between patients with MACEs and patients without MACEs where appropriate. For survival analysis, Kaplan–Meier survival curves were compared between patients with and without MACEs using log-rank statistics. Survival values are expressed together with their 95% CIs defined as survival ±1.96 * SE. Univariable and multivariable logistic regression models were used to explore the risk factors associated with MACEs. Receiver operating characteristic curves were used to find the best image parameter prediction model to predict the MACEs during hospitalization. The cutoff value was identified using the Youden index, which is the maximum of (sensitivity + specificity – 1). The Spearman rank correlation coefficient was used for linear correlation analysis. Statistical analysis was performed using SPSS 23.0 software. Results were considered significant if p < 0.05.

We finally included 241 patients [aged 61 ± 16 years (29–98 years), 115 males] with laboratory-confirmed COVID-19, 115 (47.7%) of whom had MACEs, and 126 (52.3%) of whom did not. Most of the patients are in general [108 (44.8%)] and severe [109 (45.2%)] stage; only 24 (10.0%) cases were in critical stage. There were more severe and critical patients in MACE group than those without MACE group (52.2% vs. 38.9%, 16.5% vs. 4.0%, p < 0.001). The most common symptoms at admission were fever [186 (77.2%)], cough [133 (55.2%)], and fatigue [92 (38.2%)]. Shortness of breath was more common in patients with MACEs (33% vs. 21.4%, p = 0.04). The most common comorbidities were hypertension [96 (39.8%)], diabetes [41 (17.0%)], cardiovascular disease [39 (16.2%)], and cerebrovascular disease [13 (5.4%)], but only the differences in cardiovascular disease and cerebrovascular disease were statistically significant (21.7% vs. 11.1%, p = 0.049; 8.7% vs. 2.4%, p = 0.03). In terms of laboratory findings, patients with MACEs showed lower median lymphocyte counts [median (IQR), 0.9 (0.6–1.2) ×10⁹ cells/L vs. 1.4 (1.1–1.8) ×10⁹ cells/L, p < 0.001], platelets [median (IQR),198 (140–258) ×10⁹ cells/L vs. 225 (182–275) ×10⁹ cells/L, p = 0.01], hemoglobin [median (IQR),110 (95–125) g/dL vs. 117 (106–132) g/dL, p = 0.006], ALB [median (IQR),36.4 (32.8–39.5) g/dL vs. 38.9 (36.7–41.2) g/dL, p = 0.001), and higher level of C-reactive protein [median (IQR), 21.4 (5.1–67.5) mg/mL vs. 5.0 (5.0–24.2) mg/mL, p < 0.001] and d-Dimer [median (IQR), 6.4 (1.5–20.9) μg/mL vs. 0.96 (0.43–4.29) μg/mL, p < 0.001]. The cardiac markers in MACE group were higher than those without MACE group, and the difference was statistically significant, including CK-MB [median (IQR), 1.4 (0.8–3.4) ng/mL, p = 0.005], myoglobin [median (IQR), 49.1 (29.9–171.8) μg/L vs. 33.8 (21.7–53.2) μg/L, p < 0.001], hs-TnI [median (IQR), 0.027 (<0.006–0.183) ng/mL vs. 0.006 (<0.0060–0.0065) ng/mL, p = 0.021], N-terminal pro–B-type natriuretic peptide [median (IQR), 1910 (1055–5270) μg/mL vs. 216.4 (79.1–365.4] μg/mL, p < 0.001]. Furthermore, compared with patients without MACEs, more respiratory support treatment was applied in patients with MACEs. The detailed demographic data and clinical characteristics are shown in Table 1.

According to chest CT findings (Table 2), most patients (90.9%) have bilateral pneumonia. There was no difference in GGO between the two groups, but there was more consolidation in the MACE group than in the no-MACE group. Measurement of LAD + LV was significantly larger in patients with MACEs (12.1 ± 1.5 cm vs. 11.4 ± 1.3 cm, p < 0.001), comparing with no-MACE group. The CTR >0.5 of the MACE group was more than that of the no-MACE group (58.3% vs. 31.7%, p < 0.001). In addition, pleural effusion (41.7% vs. 24.6%, p < 0.001), pericardial effusion (39.1% vs. 26.2%, p = 0.03), and bronchial wall thickening (11.3% vs. 3.2%, p = 0.01) were more likely to occur in MACE group. There was no statistical difference between the two groups in other findings such as fibrosis, reticular pattern, and bronchiectasis (all p > 0.05).

The mean follow-up period was 51 days (range, 10–83 days) for the MACE group and 53 days (range, 5–93 days) for no-MACE group. Of the 115 patients, 112 patients (97.4%) had heart failure, 9 patients (7.8%) had acute myocardial infarction, 7 patients (6.1%) had recurrent angina, and 1 patient (0.9%) had ventricular fibrillation. Thirty-two patients (27.8%) died in the MACE group during the follow-up period, much more than the no-MACE group (11 patients, 8.7%). Figure 3 shows Kaplan–Meier survival curves for all-cause mortality in the two groups, which showed patients with MACEs had substantially shorter survival (mortality rate: 27.8% vs. 8.7%, log-rank test: p < 0.001). In multivariable analysis, we found that underlying cardiovascular diseases [odds ratio (OR), 28.56; 95% CI, 3.26–250.20; p = 0.002) and NT-proBNP (OR, 1.007; 95% CI, 1.004–1.010; p < 0.001) were associated with increased odds of MACEs (Table 3). For imaging features, we also found that bronchial wall thickening would increase the odds of MACEs by 13.42 (95% CI, 1.94–93.05; p = 0.01). An increase in lymphocyte count and air bronchogram would reduce the odds of MACEs by 0.15 (95% CI, 0.03–0.87; p = 0.03) and 0.11 (95% CI, 0.01–0.78; p = 0.03), respectively.

For MACE prediction analysis (Figure 4), the areas under the curve (AUCs) of ROC were 0.84 (p < 0.001) in LAD + LV and 0.8 (p < 0.001) in CTR. The cutoff value of LAD + LV was 11.8 cm by ROC curve, yielding a sensitivity of 67.8%, a specificity of 83.3%, a positive predictive value of 60%, and a negative predictive value of 73.8%. CTR >0.5 yielded a sensitivity of 86.1% and a specificity of 69.0%. Our results showed that combining predictors (LAD + LV and CTR) was the best predictor for MACEs (AUC = 0.88, p < 0.001), which was a little higher than a single variable. The sensitivity and specificity were 82.6% and 80.2%, with a cutoff value of 12.1 cm and 0.51. Using LAD + LV (Figure 5A) and CTR > 0.5 (Figure 5B) was also helpful for classifying subjects at high risk of death. Mortality rates were significantly higher in patients with LAD + LV >11.8 cm and CTR >0.5 than in patients with LAD + LV <11.8 cm and CTR <0.5 (log-rank test: p < 0.001).

There were 32 patients with cardiac injury in our study population. The difference of minimum CT value (34.5 ± 8.3 HU vs. 44.5 ± 6.3 HU, p < 0.001) and mean CT value (42.5 ± 7.1 HU vs. 47.9 ± 6.0 HU, p ≤ 0.008) between patients with cardiac injury and patients without cardiac injury was statistically significant. The lateral wall of LV was more vulnerable compared to septum and anterior wall in our study. Plasma hs-TnI levels in patients with cardiac injury showed a moderate negative correlation with minimum CT value (R² = −0.636, p < 0.001) (Figure 6).

Cardiac injury is a common condition among patients hospitalized with COVID-19, and it is associated with a higher risk of in-hospital mortality. Chest CT plain scan can effectively screen a part of high-risk patients, so as to guide doctors to treat early and improve prognosis. This method can also save medical resources better in the current environment.