Introduction

Until May 6, 2020, coronavirus disease 2019 (COVID-19) has posed a serious threat to global public health with a total of 3,272,202 confirmed cases and 230,104 deaths documented in 212 countries [1]. Its pathogen is named the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) [2]. Except for fever, cough, and dyspnea as the major clinical presentations [3], COVID-19 patients may also develop different degrees of liver injury [4, 5]. The major mechanism of liver injury in COVID-19 patients is thought to be the binding of SARS-CoV-2 to angiotensin converting enzyme 2 (ACE2) receptor [6], which is highly expressed in bile duct cells [7], and then damages bile duct cells, thereby resulting in abnormal liver biochemical tests reflected by elevated alkaline phosphatase (ALP) and gamma-glutamyl transpeptidase (GGT) [8]. In addition, liver injury in SARS-CoV-2 infection may be caused by either systemic inflammation response or drug hepatotoxicity, which is supported by the first autopsy pathological analysis of a COVID-19 patient showing moderate microvesicular steatosis and mild lobular and portal activity in the liver tissue [9]. Multiple organ failure (MOF) is another possible cause of liver injury in COVID-19 patients, as SARS-CoV-2 can cause acute respiratory distress syndrome (ARDS) and MOF, thereby leading to hepatic ischemia, which could be worsened by the use of vasopressor medications, and hypoxia reperfusion injury in critically ill patients [10, 11]. Indeed, evidence also suggested that critically ill patients have a higher proportion of liver enzyme abnormality than patients with mild disease [8].

It is important for physicians, especially hepatologists, to appreciate the epidemiology and potential risk of liver injury in COVID-19 patients [12, 13]. However, the data regarding abnormal liver biochemical tests in COVID-19 patients are often heterogeneous among studies, and assessing the risk of liver injury in such patients remains challenging. In this study, we have systematically collected the current evidence with two major objectives: (1) to achieve more generalizable conclusions regarding the incidence of abnormal liver biochemical tests in COVID-19 patients; and (2) to explore the relationships of abnormal liver biochemical indicators with the severity and prognosis of COVID-19 patients.

Methods

This meta-analysis was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement and the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) guidelines. The MOOSE and PRISMA checklists are shown in the Supplementary Materials.

Search strategy

All published literature which reported liver biochemical tests in COVID-19 patients were identified via the PubMed, Embase, Cochrane Library, China National Knowledge Infrastructure (CNKI), VIP, and Wanfang databases. The search terms were (“2019 novel coronavirus-infected pneumonia” OR “COVID-19” OR “2019 novel coronavirus” OR “2019 novel coronavirus pneumonia” OR “severe acute respiratory syndrome coronavirus 2” OR “SARS-CoV-2” OR “novel coronavirus pneumonia”) AND (“liver” OR “hepatic”). The last retrieval date was April 27, 2020.

Study selection

There was neither publication language nor publication status restriction. All eligible studies reported the incidence and/or risk factors of abnormal liver biochemical tests in COVID-19 patients. Exclusion criteria were as follows: (1) duplicates; (2) case reports, reviews or meta-analyses, guidelines, consensus, experimental or animal studies, comments or letters, notes, and correspondences; (3) irrelevant papers; (4) absence of detailed data; (5) duplicate study population.

Data extraction

Data extraction was performed by 2 investigators. The following data were extracted from the included studies: the first author, publication year, region, source of cases, enrollment period, cases with COVID-19, age, gender, history of pre-existing liver diseases, treatments, clinical outcomes, and liver biochemical tests at admission or during hospitalization. Liver biochemical indicators analyzed included alanine aminotransferase (ALT), aspartate aminotransferase (AST), ALP, GGT, total bilirubin (TBIL), and albumin (ALB).

Study quality

The quality of cohort studies was assessed using the Newcastle–Ottawa Scale (NOS), which includes study selection (four items), comparability (two items), and exposure/outcome (three items). The highest score is 9 points, and a score of more than 6 points is considered high quality.

Definitions

The definition regarding the severity of COVID-19 patients was inconsistent among these included studies. Among them, 36 studies employed the definitions from the Chinese practice guidelines regarding diagnosis and treatment of novel coronavirus pneumonia, in which patients were divided into four subtypes (mild, moderate, severe, and critical); 4 studies employed the definitions from the American Thoracic Society guidelines, in which patients were categorized into severe and non-severe types; and the remaining 5 studies did not clearly report the definition regarding the severity of COVID-19 patients. Therefore, as for the present systematic review, the severity of COVID-19 patients was mainly dependent upon the definitions from each individual study.

The time when liver biochemical indicators were measured during hospitalization was not strictly defined in the present systematic review, because it was not strictly or consistently defined among these included studies. Generally, liver biochemical indicators at admission refer to those measured at admission or within 24 h after admission; and liver biochemical indicators during hospitalization refer to those measured at any time during hospitalization.

Statistical analysis

All statistical analyses were performed using StatsDirect statistical software version 2.8.0 (StatsDirect Ltd., Sale, Cheshire, UK), STATA version 12.0 (Stata Corp., College Station, Texas, USA), and Review Manager software version 5.3 (Cochrane collaboration, the Nordic Cochrane Centre, Copenhagen, Denmark). First, the incidence of abnormal liver biochemical tests from each study were pooled, and the pooled proportion with 95% confidence interval (CI) was calculated. Second, we calculated the risk ratios (RRs) or mean differences (MDs) with 95% CIs. The meta-analyses were performed by using a random-effect model. Heterogeneity among the studies was also assessed. I2 > 50% and/or p < 0.1 were considered to have statistically significant heterogeneity. Publication bias was assessed with Egger test. p < 0.1 was considered as a statistically significant publication bias. Subgroup analyses were performed according to the source of cases (single-center versus multiple-center), sample size (≥ 100 versus < 100), NOS (≥ 7 versus < 7), proportion of male (≥ 50% versus < 50%), and proportion of patients with pre-existing liver disease (≥ 10% versus < 10%). Meta-regression analyses employed the covariates that were the same as the strata of subgroup analyses, including the source of cases, sample size, NOS, proportion of male, and proportion of patients with pre-existing liver disease. Sensitivity analyses were conducted by sequentially excluding one study at a time. These analyses were performed to explore the sources of heterogeneity among studies.

Results

Study selection

Overall, 2281 papers were identified via the 6 databases, and 11 papers were identified via a manual search. Finally, 45 studies were included in this meta-analysis (Fig. 1) [14–58].

Fig. 1
figure 1

Flow chart of study selection

Full size image

Study characteristics

Characteristics of the included studies are summarized in Table 1. Among them, 30 studies [15, 16, 19, 21, 23, 25,25,26,27,28,30, 33,33,34,35,37, 39,40,41,43,44,45,46, 49, 50, 52,53,54,55,56] were formally published as full texts, 9 studies [18, 24, 31, 32, 38, 42, 47, 48, 51] were published in press, and the remaining 6 studies [14, 17, 20, 22, 57, 58] were preprinted. The number of COVID-19 patients ranged from 18 to 1099 among these included studies. Nearly all of these included studies (44/45) were conducted in China, and the remaining one in Singapore; 33 [14,15,16, 18, 20,21,22, 24,25,26, 28–35, 38, 39, 41, 43,44,45,46,47, 49,50,51,52, 54, 56, 58] and 12 studies [17, 19, 23, 27, 36, 37, 40, 42, 48, 53, 55, 57] were single-center and multi-center studies, respectively. Treatment and clinical outcomes of COVID-19 patients are summarized in Supplementary Table 1.

Table 1 An overview of included papers regarding COVID-19
Full size table

Study quality

The NOS score ranged between 3 and 8 points. Eight studies [15, 16, 21, 23, 29, 36, 40, 50] were considered to be of high quality (Supplementary Table 2).

Incidence

The results of meta-analyses regarding the incidence of abnormal liver biochemical tests in COVID-19 patients are shown in Table 2.

Table 2 Incidence of abnormal liver biochemical indicator: results of meta-analyses
Full size table

Overall analyses

The pooled incidence of any abnormal liver biochemical indicator at admission was 27.2% (95% CI 19–36.3%). The pooled incidence of abnormal ALT, AST, ALP, GGT, TBIL, and ALB at admission was 20.4% (95% CI 16.8–24.3%), 21.8% (95% CI 17.6–26.3%), 4.7% (95% CI 1.8–8.9%), 35.8% (95% CI 17.8–56.1%), 8.8% (95% CI 5.5–12.8%), and 39.8% (95% CI 30.6–49.5%), respectively.

The pooled incidence of any abnormal liver biochemical indicator during hospitalization was 36% (95% CI 12.3–57.1%). The pooled incidence of abnormal ALT, AST, and TBIL during hospitalization was 38.4% (95% CI 24.2–53.7%), 28.1% (95% CI 15.9–42.2%), and 23.2% (95% CI 6–64.2%), respectively.

Subgroup analyses

The results of subgroup analyses are summarized in Supplementary Table 3. The heterogeneity remained statistically significant in all subgroup analyses.

Meta-regression analyses

The results of meta-regression analyses are summarized in Supplementary Table 4. As for any abnormal liver biochemical indicator observed at admission, meta-regression analyses indicated that the source of cases (p = 0.000) might be a potential source of heterogeneity. As for abnormal ALT observed at admission, meta-regression analyses indicated that the sample size (p = 0.019) might be the potential source of heterogeneity. As for any abnormal liver biochemical indicator observed during hospitalization, meta-regression analyses indicated that the proportion of patients with pre-existing liver disease (p = 0.028) might be a potential source of heterogeneity. However, no source of heterogeneity could be identified for other liver biochemical indicators observed at admission or during hospitalization.

Sensitivity analyses

Sensitivity analyses did not identify any study as the potential source of heterogeneity.

Risk factors

Non-severe versus severe according to the American Thoracic Society guideline

Two studies [17, 19] compared the incidence of abnormal ALT, AST, and TBIL at admission between severe and non-severe patients. Meta-analyses demonstrated that severe patients had a significantly higher incidence of abnormal AST at admission (RR = 2.91, 95% CI 1.36–6.22; p = 0.006). The incidence of abnormal ALT (RR = 2.32, 95% CI 0.78–6.88; p = 0.13) and TBIL (RR = 1.95, 95% CI 0.68–5.58; p = 0.21) at admission were not significantly different between severe and non-severe patients.

Mild and moderate versus severe and critical according to the Chinese practice guideline

Only one study [38] compared the incidence of any abnormal liver biochemical indicator at admission between mild and moderate patients versus severe and critical patients, and demonstrated that severe and critical patients had a significantly higher incidence of any abnormal liver biochemical indicator at admission (RR = 2.78, 95% CI 1.28–6.06; p = 0.01). Three studies [49, 51, 55] compared the incidence of abnormal ALT at admission between mild and moderate patients versus severe and critical patients, and demonstrated that severe and critical patients had a significantly higher incidence of abnormal ALT at admission (RR = 3.03, 95% CI 1.76–5.23; p < 0.0001). Four studies [35, 49, 51, 55] compared the incidence of abnormal AST at admission between mild and moderate patients versus severe and critical patients, and demonstrated that severe and critical patients had a significantly higher incidence of abnormal AST (RR = 3.84, 95% CI 2.53–5.82; p < 0.00001) at admission. Two studies [49, 51] compared the incidence of abnormal TBIL at admission between mild and moderate patients versus severe and critical patients, and demonstrated that severe and critical patients had a significantly higher incidence of abnormal TBIL at admission (RR = 3.16, 95% CI 1.24–8.09; p = 0.02).

Eight studies [16, 20, 30, 35, 49, 51, 55, 57] compared the ALT level at admission between mild and moderate patients versus severe and critical patients. Meta-analyses demonstrated that ALT level was significantly higher in severe and critical patients than in mild and moderate patients (MD = 7.64, 95% CI 1.94–13.35; p = 0.009).

Seven studies [16, 20, 30, 35, 49, 51, 55] compared the AST level at admission between mild and moderate patients versus severe and critical patients. Meta-analyses demonstrated that AST level was significantly higher in severe and critical patients than in mild and moderate patients (MD = 13.20, 95% CI 7.57–18.82; p < 0.00001). Seven studies [20, 30, 35, 49, 51, 55, 57] compared the TBIL level at admission between mild and moderate patients versus severe and critical patients. Meta-analyses demonstrated that TBIL level was lower in severe and critical patients than in mild and moderate patients (MD = − 0.22, 95% CI − 5.08 to 4.63; p = 0.93). Five studies [16, 30, 35, 55, 57] compared the ALB level at admission between mild and moderate patients versus severe and critical patients. Meta-analyses demonstrated that ALB level was significantly lower in severe and critical patients than in mild and moderate patients (MD = − 4.84, 95% CI − 6.68 to − 3.00; p < 0.00001).

Moderate versus severe and critical according to the Chinese practice guideline

Only one study [18] compared the incidence of any abnormal liver biochemical indicator at admission between moderate patients versus severe and critical patients, and demonstrated that severe and critical patients had a significantly higher incidence of any abnormal liver biochemical indicator at admission (RR = 1.91, 95% CI 1.17–3.1; p = 0.009).

Three studies [15, 34, 52] compared the ALT, AST, TBIL and ALB levels at admission between moderate patients and severe and critical patients. Meta-analyses demonstrated that the ALT (MD = 9.76, 95% CI 4.64–14.89; p = 0.0002), AST (MD = 6.32, 95% CI 2.90–9.74; p = 0.0003) and TBIL (MD = 3.14, 95% CI 1.14–5.14; p = 0.002) levels were significantly higher in severe and critical patients than in moderate patients, and the ALB level was significantly lower in severe and critical patients (MD = − 7.30, 95% CI − 8.69 to − 5.90; p < 0.00001).

Moderate versus severe according to the Chinese practice guideline

Five studies [26, 28, 29, 45, 54] compared the ALT and AST levels at admission between moderate and severe patients. Meta-analyses demonstrated that the ALT (MD = 11.99, 95% CI − 3.59 to 27.57; p = 0.13) and AST (MD = 10.30, 95% CI 0.11–20.49; p = 0.05) levels were higher in severe patients than in moderate patients. Three studies [26, 28, 54] compared the ALB level at admission between moderate and severe patients. Meta-analyses demonstrated that ALB level was significantly lower in severe patients than in moderate patients (MD = -4.62, 95% CI − 8.12 to − 1.13; p = 0.01). Two studies [26, 45] compared the TBIL level at admission between moderate and severe patients. Meta-analyses demonstrated that the TBIL level was lower in severe patients than in moderate patients (MD = − 0.12, 95% CI − 0.87 to 0.62; p = 0.75).

At admission versus during hospitalization

Two studies [32, 47] compared the ALT level detected at admission versus during hospitalization, and demonstrated that the ALT level detected during hospitalization was significantly higher than that obtained at admission (MD = 20.30, 95% CI 16.51–24.06; p < 0.00001).

The first week after admission versus the second week after admission

Only one study [38] compared the ALT level at the first week after admission versus the second week after admission, and demonstrated that ALT level detected at the second week after admission was significantly higher than that detected at the first week after admission (MD = 130.75, 95% CI 116.14–145.36; p < 0.00001).

Number of drug products combined

Only one study [38] compared the proportion of drug products ≥ 3 between patients with liver injury and those with normal liver biochemistry, and demonstrated that the proportion of drug products ≥ 3 was significantly higher in patients with liver injury than in those with normal liver biochemistry (RR = 9.00, 95% CI 1.28–63.26; p = 0.03).

Effect of liver biochemical indicators on prognosis of COVID-19 patients

Survivors versus non-survivors

Three studies [43, 44, 50] compared the incidence of any abnormal liver biochemical indicator between survivors versus non-survivors. Meta-analyses demonstrated that non-survivors had a significantly higher incidence of any abnormal liver biochemical indicator (RR = 1.34, 95% CI 1.02–1.77; p = 0.04).

With versus without composite endpoint

Two studies [19, 50] compared the incidence of abnormal ALT and AST between patients who achieved the composite endpoint versus those who did not achieve the composite endpoint. Both of them employed the same composite endpoint defined as admission to the ICU, mechanical ventilation, or death. Meta-analyses demonstrated that patients who achieved the composite endpoint had a significantly higher incidence of abnormal ALT (RR = 1.96, 95% CI 1.54–2.49; p < 0.00001) and AST (RR = 2.30, 95% CI 1.81–2.92; p < 0.00001).

Discussion

Our study suggested that the pooled incidence of any abnormal liver biochemical indicator detected during hospitalization seemed to be higher than that detected at admission (36% versus 27.2%). Except for the disease progression of COVID-19 during hospitalization, this might be partly due to the toxicity of drugs used during hospitalization. Patients are often given empirical antiviral therapy for COVID-19 after admission, of which some can be potentially hepatotoxic. At the time of this writing, recent studies have reported that lopinavir might be potentially effective against SARS-CoV-2 [59], but lopinavir/ritonavir is mainly metabolized by cytochrome P 450 3A4 (CYP3A4) enzymes in the liver, which is likely to cause elevated serum transaminases [60]. In addition, some patients may be treated with antipyretic agents during hospitalization, of which most contain acetaminophen that can be hepatotoxic in high doses and/or combination with other drugs and even cause liver failure [61].

Among the abnormal liver biochemical indicators observed at admission, abnormal ALB (39.8%) was the most frequent, followed by abnormal GGT (35.8%), AST (21.8%), ALT (20.4%), TBIL (8.8%), and ALP (4.7%). Decreased ALB level is usually considered to indicate that the synthetic function of the liver be damaged to some extent. However, a decline of ALB level may be related to the disease severity of COVID-19, as COVID-19 can cause pulmonary exudation, thereby leading to abnormal ALB distribution, and an insufficient intake of nutrients or impairment of normal utilization/metabolism of nutrients may also decrease the ALB level in COVID-19 patients. SARS-CoV-2 is prone to damage bile duct cells where the ACE2 is highly expressed [7]. Thus, abnormal GGT and ALP levels should have been more common. However, in the settings of liver injury, the de-differentiation and proliferation of ACE2-expressing bile duct epithelial cells are involved in liver tissue repair, and some newborn hepatocytes retain the characteristics of ACE2 expression and may be susceptible to SARS-CoV-2 [62], thereby affecting hepatocytes and presenting with abnormal AST and ALT levels. Both ALP and GGT are considered as cholangiocyte-related enzymes, but the pooled incidence of abnormal ALP seems to be remarkably higher than that of abnormal GGT (35.8% versus 4.7%). This counter-intuitive phenomenon may be attributed to a difference in the distribution of ALP and GGT. ALP is present in bile duct, bone, intestine, kidney, and placenta, while GGT is widely distributed in the cell membranes of many tissues, such as bile duct, kidney, pancreas, gallbladder, spleen, heart, brain, and seminal vesicle. Thus, GGT, as an indicator of bile duct injury, may be less sensitive than ALP.

Severe and critical COVID-19 patients and non-survivors are more prone to have abnormal liver biochemistry. This finding might be explained by several points, as follows. First, inflammatory factor storm is suspected to be associated with abnormal liver biochemistry in severe and critical COVID-19 patients. It has been recognized that the occurrence of MOF is mainly associated with the sudden initiation of an inflammatory storm in the critically ill patients [63]. The release of numerous inflammatory cytokines induces ARDS and systemic inflammatory response syndrome (SIRS) and subsequently causes hypoxia in the body, thereby leading to an injury in lung, liver, myocardium, and kidney. This may be aggravated by the use of vasopressor drugs to maintain blood pressure in an ICU setting. Second, hepatic ischemia and hypoxia reperfusion dysfunction may be one of the main mechanisms of liver injury in severe and critical COVID-19 patients. COVID-19 patients often have varied degrees of hypoxemia, of whom more than 40% need to receive oxygen therapy [19]. COVID-19-related complications include ARDS, SIRS, and MOF, which may cause hepatic ischemia and hypoxia reperfusion dysfunction. Both in vivo and in vitro models of liver ischemia and hypoxia suggested that liver cell death and inflammatory cell infiltration could be caused by ischemia and hypoxia [11]. Meanwhile, oxygen reduction and lipid accumulation in liver tissue during shock and hypoxic conditions may further promote the release of multiple inflammatory factors and then lead to liver injury [64]. Third, drug toxicity may contribute to liver damage in severe and critical COVID-19 patients. Compared with patients experiencing a mild or moderate clinical course, severe and critical patients require longer duration of antiviral therapy and multiple drugs combined. Our findings found that the number of drugs products ≥ 3 might be related to liver injury. Additionally, it has been reported that antiviral medications (lopinavir/ritonavir, arbidol, hydroxychloroquine), antipyretics (acetaminophen), antibiotics (macrolides, quinolones), and traditional Chinese medicine can cause liver damage [60, 65,66,67]. Also, some critical patients would be treated with steroids, which are mainly metabolized in the liver and could cause mild hepatotoxicity. In the setting of steroids combined with HIV protease inhibitors (such as lopinavir), the risk of liver damage could be further increased [68]. Notably, at the beginning of the COVID-19 crisis, in the complete absence of known antiviral/disease modulating medications, many drugs and combinations of drugs, which would have been previously considered for use only in a controlled experimental setting, may have been used empirically in clinical practice. Fourth, our subgroup analysis suggested that the incidence of any abnormal liver biochemical indicator at admission seemed to be higher in the subgroup where the proportion of pre-existing liver disease was ≥ 10% than in the subgroup where the proportion of re-existing liver disease was < 10%. This finding suggests that COVID-19 patients with pre-existing liver diseases may be more prone to have abnormal liver biochemical indicators and that monitoring liver function should be more intensive in such patients. Notably, this may not be a direct consequence of pre-existing liver diseases, but related to the dysfunction of innate immune response against the virus [69].

Our study has several limitations. First, the heterogeneity remained statistically significant in most of our meta-analyses. Despite this, we performed subgroup analyses, meta-regression analyses, and sensitivity analyses, but the potential source of heterogeneity could not be clearly identified. Indeed, among the proportion meta-analyses, a statistically significant heterogeneity is often unavoidable [70, 5]. Second, the follow-up duration was different among these included studies. Third, all but one included study was from China and even that study was also from its neighboring country. Thus, the present findings might be more appropriate for the Chinese population or Asian population and it remains to be seen whether our findings can be generalized to European or American populations. In the future, it will be interesting to conduct a similar systematic review with papers that are likely to be forthcoming from Europe and America. However, at the time of writing this meta-analysis, the papers from Europe and America have yet to appear in the literature. Fourth, most of these included studies were retrospective, which might cause a recall bias, especially about data entry. Fifth, abnormal liver biochemical test, rather than liver injury, was assessed, because the definition of liver injury was unclear or inconsistent among these included studies.

In conclusion, abnormal liver biochemistry, primarily characterized as decreased ALB and elevated GGT, AST, and ALT, is common in COVID-19 patients. Abnormal liver biochemical indicators are closely related to the severity and prognosis of COVID-19 patients. Additionally, a higher incidence of abnormal liver biochemistry in COVID-19 patients during hospitalization warrants that liver biochemical tests should be closely monitored and timely measures should be taken. In future, it is necessary to extrapolate these findings to patients who have advanced chronic liver diseases or liver cirrhosis, because there is a potential chance for co-incidental COVID-19 infection to result in hepatic decompensation in patients with little hepatic reserve.