The study was approved by the Medical Ethics Committee of Wuhan Central Hospital, Tongji Medical College, Huazhong University of Science and Technology (No. 202074). Before enrollment in this study, verbal consent was obtained from the patients or an accompanying relative if the patients were unable to provide consent.

The cohort in this study consisted of patients from two designated medical centers affiliated with the Central Hospital of Wuhan, University of Science and Technology (Wuhan, China). These centers were the closest healthcare centers to the area where the first case of COVID-19 was reported. Patients hospitalized from January 1 to March 15, 2020 who were determined to be COVID-19-positive according to the WHO interim guidelines were evaluated. A positive case of COVID-19 was defined based on positive results from a real-time reverse-transcription polymerase chain reaction (RT-PCR) test or specific antibodies against SARS-CoV-2 in the serum. A total of 1046 patients treated up to April 15, 2020 were characterized and assessed. The following patients were excluded: 179 patients with negative COVID-19 results in serial RT-PCR, 13 patients with other pathogen pneumonia unrelated to COVID-19, 19 patients that lacked follow-up clinical records and radiographic data, and five patients with poor quality images of the liver. Therefore, 830 patients with COVID-19 who underwent radiological assessment were eligible for inclusion. Liver biochemistry abnormality (LBA) was defined as ALT, AST, GGT, signifying the liver injury, whereas total bilirubin levels that exceeded the upper limit of the normal value, signifying the liver function. LI was defined as ALT or AST levels at three times the upper limit of the normal values. LI was also defined as total bilirubin levels that exceeded twice the upper limit of the normal value, in accordance with expert consensus from the Chinese Society of Hepatology, Chinese Medical Association [15]. The updated Roussel Uclaf Causality Assessment Method (RUCAM) was used to determine the causality between drugs application and liver injury. The final scores correspond to the following causality levels: =0 points, excluded causality; 1–2, unlikely; 3–5, possible; 6–8, probable; and =9, highly probable [16].

The medical history, laboratory tests, and computed tomography (CT) data for all patients were collected from the institutional digital medical records database. The data for all patients included the basic features of patients (age, gender), main symptoms (fever, cough, diarrhea, anorexia, weakness), and comorbidities (hypertension, diabetes, hyperlipidemia, cardiac or cerebrovascular disease, malignancy, chronic kidney disease, chronic pulmonary disease, surgery, liver fibrosis/cirrhosis, and HBV). Lymphocyte count, lymphocyte proportion, C-reactive protein (CRP), d-dimer, albumin, ALT, AST, AST/ALT ratio, GGT, total bilirubin, triglyceride, and cholesterol test results on hospital admission were also recorded. Confirmed clinical cases were classified based on the final diagnosis upon discharge using the Chinese Guidelines for COVID-19 (Trial version 7.0) [17]. Other related clinical diagnostic criteria were in accordance with the preliminary diagnosis and treatment protocols from the National Health Commission of the People’s Republic of China (Trial version 7.0). To assess hepatic attenuation in patients, 3-mm sections on chest CT images of the livers of the patients were independently reviewed by two trained radiologists who had no access to other parameters. Visual assessment of steatosis in the right hepatic lobe was determined with a ?ve-point scale described in a previous study by Lee et al. [18]. Liver attenuation in a 300-mm2 rectangular area in the right lobe was examined (Fig. 1). Disagreements between the two reviewers were resolved by a third reviewer with proper consultation.

Fig. 1.

A 39-year-old man with COVID-19 pneumonia. Changes in liver attenuation on serial chest CT examinations. (A) a section of the liver shows hypoattenuation (mean attenuation: 34.4 HU) in the liver parenchyma on initial chest CT scan (obtained on day 4 after the onset of symptoms); (B) a section of the liver shows an increase in liver attenuation (mean attenuation: 48.8 HU) on the second chest CT (day 10 after the onset of symptoms); (C) a section of the liver shows continued recovery in liver attenuation (mean attenuation: 54.0 HU) on the third chest CT scan (day 15 after the onset of symptoms); (D) initial chest CT scan shows multifocal patchy ground-glass opacities (GGO) in two lungs; (E) the second chest CT image shows dissipation of GGO in the left lower lobe and a mild enlargement of GGO in the right lower lobe; (F) the third chest CT image shows demonstrative absorption.

(0.81MB).

Continuous variables were expressed as the median with the interquartile range (IQR) and compared using the Kruskal–Wallis and Mann–Whitney U tests where appropriate. Categorical variables were presented as frequency (%), and proportions in categorical variables were compared using the ?2 test or Fisher’s exact test. Univariable and multivariable logistic regression models were used to explore the risk factors associated with LBA and LI. Since lymphocyte proportion, lymphocyte count, and CRP levels can contribute to systemic inflammation, multivariable analysis was used for the calculation of lymphocyte proportion. Receiver operating characteristic (ROC) analysis was used to determine the predictive values of LBA and LI. A two-sided a of less than 0.05 was considered statistically significant. Statistical analyses were performed using SPSS v. 25.0 (IBM, Armonk, NY) and MedCalc v. 17.6 (MedCalc Software BVBA; https://www.medcalc.org).

As shown in Table 1, among 830 COVID-19 patients, the median ages of severe and critical patients were 62 and 70 years old, respectively, and over 60% of the patients were men. By March 15, 2020, 1043 patients had been discharged from the designated medical centers after they had fully recovered and met the clinical criteria for COVID-19 recovery. By April 10, 2020, 196 patients were either still hospitalized or the tests for SARS-CoV-2 nucleic acids or serum-specific antibodies for SARS-CoV-2 were not confirmed yet or were negative. Key information was missing in the medical records of 17 inpatients. Therefore, the data for 830 inpatients were included in the final analysis in this study. Notably, two patients eventually developed sub-acute liver failure while they were hospitalized; one of patients was an 84-year-old woman with pancreatic ampullary carcinoma and the other was a 36-year-old man with cholangiocarcinoma. Both patients accepted endoscopic retrograde biliary drainage to improve jaundice while undergoing treatment. Out of the entire cohort, 98 (11.8%) patients died during hospitalization and 732 were discharged. The median age was 51 years old and 422 (50.8%) patients were over 50 years old. Two common gastrointestinal symptoms observed in the patients were diarrhea (1.6%) and anorexia (0.7%), with no significant differences among four groups of patients categorized according to disease severity (mild, moderate, severe, and critical). Clinically, hypertension, cardiac or cerebrovascular disease, and diabetes were the three most frequently observed chronic diseases. Twenty-four patients who had liver fibrosis/cirrhosis associated with conditions including chronic HBV, schistosome, NAFLD, and excessive alcohol intake. Eighteen patients were HBV-positive. The data suggested that there were significant differences in gender, age, fever, and comorbidities between the severely group and mild group (p? d-dimer values was also observed in severe and critically ill patients than in patients with standard symptoms.

As shown in Table 2, 227 out of 830 patients (27.3%) had different degrees of liver dysfunction on admission, specifically, mild LBA (n?=?195) and hepatic impairment (n?=?32). Among these patients, 63.9% were men. Among 227 patients, 41.9% (95/227), 48.5% (110/227), 50.7% (115/227) and 21.1% (48/227) were abnormal in serum ALT, AST, GGT, and total bilirubin, respectively. Approximately 16.3% (37/227) of the patients showed an abnormal increase in all three serum liver enzymes tested (ALT, AST, GGT). The median AST/ALT ratios, which showed a gradual decline in the groups according to the degree of liver dysfunction, were 1.16, 0.89, and 0.68 in the non-LBA, mild LBA, and LI subgroups, respectively. Compared with the non-LBA subgroup, the LBA subgroup showed lower levels of CRP, d-dimer, and aminotransferases. Moreover, the lymphocyte count and lymphocyte proportion were significantly lower in the LBA subgroup. No clear statistical difference was found between the mild LBA and LI subgroups as well. No significant differences were observed between the mild LBA and LI subgroups in terms of chronic comorbidities such as hypertension, diabetes, cardiac disease, chronic renal disease, malignancy, chronic HBV, and liver fibrosis/cirrhosis. Abnormalities in AST, GGT, and albumin were more frequently observed in severely ill patients than in non-severe patients; however, this difference was not apparent in ALT and total bilirubin. 32.6% (74/227) of the LBA patients had the updated RUCAM score >3, whereas the non-LBA patients had a slight lower at a rate of 24.2% (146/603) (P?=?0.047).

In univariable analysis, the odds of LBA development in patients were associated with the presence of hepatic steatosis, liver fibrosis/cirrhosis, and triglyceride levels higher than 1.7?mol/L (Table 3). Lymphopenia, elevated ALT/AST, and elevated CRP were also highly associated with LBA (Table 3). However, hepatic steatosis, HBV infection, and liver fibrosis/cirrhosis were not related to LI. The multivariable logistic regression model was developed with the data for all 830 patients with complete data for all variables (603 non-LBA, 195 mild-LBA, and 32 LI). We found that hepatic steatosis, lymphocyte proportion less than 20%, CRP index higher that 1?mg/dL, and AST/ALT ratio greater than 1 on admission were highly associated with increased odds of LBA (Table 3). Furthermore, lymphocyte proportion (<20%), CRP (>1?mg/dL), AST/ALT (ratio >1), and triglyceride level (>1.7?mol/L) were highly correlated with increased odds of LI (Table 4). These findings suggested that a systemic inflammation response in the patients was the primary reason for LI at admission, followed by the high level of triglyceride (Fig. 2). ROC analysis indicated the area under the curves (AUCs) for predicting LBA and LI in the multivariable logistic regression models were 0.729 (95% confidence interval [CI] 0.697-0.759) and 0.765 (95% CI 0.735-0.793). The results also suggested that the sensitivity and specificity of the model were relatively moderate (Sensitivity: LBA, 76.2%, LI, 84.8%; Specificity: LBA, 58.2%, LI, 54.7%) (Fig. 3).

Fig. 2.

Mechanism of liver injury in patients with coronavirus disease 2019 (COVID-19).

(0.47MB).

Fig. 3.

(A) ROC curve for multivariable logistic regression model in predicting LBA with the AUC. (B) ROC curve analysis of multivariable logistic regression model in predicting LI.

(0.21MB).

At hospital admission, 94 patients (11.3%, 94/830) were diagnosed with hepatic steatosis. Among them, 50 patients had normal values for the liver function metrics while 44 patients had elevated values. Thus, hepatic steatosis was more frequently observed in LBA patients (19.4%, 44/227) than in non-LBA patients (8.3%, 50/603). Indeed, hepatic steatosis was diagnosed based on the observation of CT attenuation in the liver parenchyma, which corresponds to CT liver attenuation and liver-to-spleen attenuation ratio within the ranges of 38.2?±?10.4 HU and 0.70?±?0.19, respectively. Among the patients with hepatic steatosis, 21, 21, 28, and 24 patients were assessed with hepatic steatosis scores of 2, 3, 4, and 5, respectively. There was no significant statistical difference in hepatic steatosis between severe and non-severe patients (CTL: mean 39.6?±?8.7 HU vs 37.9?±?10.8 HU, p?=?0.604; CTL/s: mean 0.72?±?0.16 vs 0.69?±?0.20, p?=?0.598). However, 27 (61.4%, 27/44) patients with LBA showed an obvious recovery in liver hypoattenuation on follow-up chest CT scans during hospitalization, evidenced by marked improvements in CT attenuation of the liver with therapeutic treatment (Fig. 1). Due to the lack of CT scans of the liver for these patients prior to hospitalization, it is not practical or feasible to compare the state of hepatic CT attenuation for these patients pre- and during hospitalization.

The present study admittedly has several limitations. First, during the COVID-19 outbreak, asymptomatic patients were not hospitalized, which can cause biases of analysis in clinical observations. Second, the patients were only recruited from two designated centers in Wuhan. The sample size of the cohort was relatively small; therefore, it would be better to include patients from other hospitals in Wuhan, other cities in China, and even other countries across the world to create a large dataset. Third, all data were retrieved from the digital medical records during hospitalization but the baseline results for the laboratory tests, liver CT, and medication history of COVID-19 patients before hospitalization were not included or available. Finally, there were some side effects from chest CT imaging on the detection of liver attenuation. Since chest CT imaging using a low-dose scan protocol can potentially affect the liver attenuation measurements, we used an iterative reconstruction technique in an attempt to improve the quality of the images [34].