Longitudinal analysis of the utility of liver biochemistry in hospitalised COVID-19 patients as prognostic markers

Background: COVID-19, the clinical syndrome caused by infection with SARS-CoV-2, has been associated with deranged liver biochemistry in studies from China, Italy and the USA. However, the clinical utility of liver biochemistry as a prognostic marker of outcome for COVID-19 is currently debated. Methods: We extracted routinely collected clinical data from a large teaching hospital in the UK, matching 585 hospitalised SARS-CoV-2 RT-PCR-positive patients to 1165 hospitalised SARS-CoV-2 RT-PCR-negative patients for age, gender, ethnicity and pre-existing comorbidities. Liver biochemistry was compared between groups over time to determine whether derangement was associated with outcome. Results: 26.8% (157/585) of COVID-19 patients died, compared to 11.9% (139/1165) in the non-COVID-19 group (p<0.001). At presentation, a significantly higher proportion of the COVID-19 group had elevated alanine aminotransferase (20.7% vs. 14.6%, p=0.004) and hypoalbuminaemia (58.7% vs. 35.0%, p<0.001), compared to the non-COVID-19 group. Within the COVID-19 group, those with hypoalbuminaemia at presentation had 1.83-fold increased hazards of death compared to those with normal albumin (adjusted hazard ratio [HR] 1.83, 95% CI 1.25-2.67), whilst the hazard of death was ~4-fold higher in those aged [≥]75 years (adjusted HR 3.96, 95% CI 2.59-6.04) and ~3-fold higher in those with pre-existing liver disease (adjusted HR 3.37, 95% CI 1.58-7.16). In the COVID-19 group, alkaline phosphatase increased (R=0.192, p<0.0001) and albumin declined (R=-0.123, p=0.0004) over time in patients who died. We did not find a significant association between other liver biochemistry and death. Conclusion: In this UK population, liver biochemistry is commonly deranged in patients with COVID-19 but only baseline low albumin and a rising alkaline phosphatase over time are prognostic markers for death.


Introduction
Over 28 million confirmed cases of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) infection and 922,252 deaths have been reported globally as of 14 th September 2020; and the UK is one of the worst affected countries with 368,508 confirmed cases and 41,628 deaths reported (1). The clinical syndrome caused by SARS-CoV-2 is COVID-19, which primarily affects the respiratory system, but other organs, including the heart, gastrointestinal tract and liver may be affected, and a systemic sepsis syndrome may develop (2).
Albumin is a non-specific marker of liver function, and has been less consistently assessed; it is typically reported as one parameter of patients' baseline characteristics, with limited investigation of its impact as a prognostic marker. A recent meta-analysis of 20 retrospective studies from China revealed that patients with severe COVID-19 had lower albumin level compared to mild cases, but there was significant heterogeneity between studies (11).
Another meta-analysis showed that hypoalbuminaemia could be included in prognostic machine learning models to predict severe COVID-19 or mortality (12).
Several studies have investigated potential associations between liver biochemistry and death in COVID-19 patients (6-8), or included liver biochemistry in the development of predictive models (13-16). A report from Italy showed that ALP >150 U/L at admission (without adjusting for relevant confounders) was associated with clinical deterioration in 292 COVID-19 patients (7), and another study from the USA reported that peak ALT >5 times upper limit of normal (ULN) during admission was associated with death in a cohort of 2,273 6 COVID-19 patients (8). However, another USA study reported that elevations in ALT and AST elevation on admission were associated with length of stay, ICU admission and intubation but not death (6). Studies from Wuhan, China did not find associations of ALT (13,14) or AST (14) elevation on admission with death in COVID-19 patients, whilst several studies have reported associations of elevated total bilirubin (14) and low albumin on admission (14-16) with risk of death have been reported. Given these variable associations between liver biochemistry and COVID-19 outcomes, the prognostic value of liver biochemistry derangement in COVID-19 needs further evaluation.
Having established a clinical data pipeline through the National Institute for Health Research (NIHR) Health Informatics Collaborative (HIC) (17, 18), our tertiary referral hospital in the UK is strongly placed to undertake analyses using electronic health data from hospitalised patients. Using this resource, we aimed to determine the prevalence of deranged liver biochemistry at baseline and over the disease course in COVID-19 patients, with comparison to a matched group of non-COVID-19 patients admitted during the same period.
We also aimed to determine whether baseline liver biochemistry derangement was associated with risk of death in COVID-19 patients, and to compare longitudinal changes in liver biochemistries between COVID-19 patients who died and who survived.

Data collection
We used routinely collected clinical data from Oxford University Hospitals (OUH) NHS Foundation Trust, a large teaching hospital trust in the South East of the UK, with ~1000 inpatient beds. The data is collected by the local NIHR HIC team in Oxford, being drawn automatically from operational systems into a data warehouse and linked to produce a comprehensive record for each patient, as previously described (17). All of the data used for this studied were provided in anonymised form by OUH NHS Foundation Trust, with the prior approval of the Trust Information Governance Team, following the satisfactory completion of a Data Protection Impact Assessment. The management of the dataset is governed by the NIHR HIC Data Sharing Framework. Research Ethics Committee approval was obtained for the database and associated activity: IRAS ID 174658; REC reference 15/SC/0523.
The data extracted for this study included detailed information on demographics, body mass index (BMI), emergency admissions, blood test results, diagnostic codes, procedures, intensive care unit admission, prescriptions, medicines administration, and discharge All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this this version posted September 18, 2020. . https://doi.org/10.1101/2020 destination/outcome for all patients admitted to OUH between 1 st January 2020 and 21 st August 2020.

Inclusion and exclusion criteria
To select eligible data for adults with/without COVID-19, the inclusion criteria were: (a) at least one RT-PCR nose/throat swab having been undertaken (COVID-19 patients were

Definitions
We defined baseline as the date of the first positive SARS-CoV-2 RT-PCR test for a COVID-19 patient and the date of the first negative test for a patient without COVID-19. The end of follow-up was defined as the date of last available electronic patient record data for a patient.
The normal ranges set by the hospital biochemistry lab are: ALT: 10-45 IU/L; ALP: 30-130 IU/L; BR: 0-21 umol/L; Albumin: 32-50 g/L. The reference ranges for other blood tests are provided in Table S1. We defined baseline liver biochemistry (ALT, ALP, BR, or albumin) as the liver biochemistry measured within 7 days of SARS-CoV-2 RT-PCR test and baseline derangement as at least one abnormal result at this time point. We defined peak/nadir liver biochemistry derangement as at least one abnormal value at any point during follow-up. We defined liver biochemistry recovery as normalisation following derangement. The primary outcome was death during follow-up, and secondary outcomes included ICU admission and invasive ventilation.
Pre-existing comorbidities were defined by a historical diagnosis of a disease before baseline, and diagnosis codes retrieved were provided in Table S2. A full list of prescribed drugs searched for data extraction is provided in Table S3.
All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Statistical analysis
Statistical analysis was performed using R version 4.0.2. All significance tests performed were two-sided. p values <0.05 were deemed statistically significant.

Propensity Score Matching
We selected eligible subjects, and conducted propensity score matching process to ensure the COVID-19 group was comparable to the non-COVID-19 group in terms of demographics and pre-existing conditions. We used the following variables to calculate propensity scores: age, gender, ethnicity, and pre-existing comorbidities (liver disease, diabetes mellitus (DM), hypertension (HTN), coronary heart disease (CHD), chronic kidney disease (CKD), and cancer). We performed propensity score matching using the package MatchIt, with the nearest-neighbour method applied, the matching ratio and calliper size set as 1:2 and 0.1 respectively, and without replacement.

Comparison of COVID-19 and matched non-COVID-19 patients
For continuous variables, we calculated median and interquartile range (IQR), or mean and standard deviation (SD), and used Wilcoxon test or t-test for comparison. For categorical variables, we computed number and percentage, and used chi-square or Fisher's exact test for comparison. We used the Shapiro-Wilk test and graphical methods for normality check.
Investigation on whether liver biochemistry predict outcomes in COVID-19 patients We compared the presence of clinical outcomes in the COVID-19 and non-COVID-19 groups. We also performed Kaplan-Meier (K-M) analysis to compare the survival probability over time in COVID-19 and non-COVID-19 groups. Within the COVID-19 and non-COVID-19 groups, we compared demographics, BMI, comorbidities, baseline and peak/nadir liver biochemistry between those who died vs. survived. We then performed K-M analysis to compare the survival probabilities over time between subgroups with and without deranged baseline liver biochemistry. We used univariate and multivariate Cox proportional-hazards models to investigate whether liver biochemistry predicted death, reporting hazard ratios (HRs) and 95% confidential intervals (CIs). To investigate associations of additional patient characteristics with risk of death, and the robustness of HRs to adjustment for additional confounders, we performed sensitivity/subset analysis whereby associations were investigated in a subset of patients who were not missing data for confounders.

Longitudinal analysis of liver biochemistry in COVID-19 patients
We compared liver biochemistry between COVID-19 group and non-COVID-19 group at each time point (to examine difference between groups), as well as investigating liver biochemistry changes over time by comparing liver biochemistry at subsequent time points to their baseline within each group (to examine longitudinal changes). Within the COVID-19 group, we performed longitudinal analysis of liver biochemistry in subgroups of patients who died and survived by fitting linear regression lines with 95% CIs, and reporting Pearson's correlation coefficients and linear regression significance.

Identification, demographics, and outcomes of COVID-19 compared to non-COVID-19 groups
We identified 6311 eligible patients (585 adults with SARS-CoV-2 infection and 5726 without) according to the inclusion/exclusion criteria ( Figure S1). Based on our 585 COVID-19-positive patients, we matched a cohort of 1165 COVID-19-negative patients. After matching, there were no significant differences in demographics and pre-existing comorbidities between the two groups ( All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Over a median follow-up of 58 days vs. 50 days, 26.8% (157/585) of COVID-19 patients died compared to 11.9% (139/1165) in the non-COVID-19 group (p<0.001), respectively. The COVID-19 group had a higher rate of ICU admission during hospitalisation (12.1% vs. 4.3%, p<0.001), and a higher rate of invasive ventilation use in ICU (8% vs. 2.5%, p<0.001) ( Table   1). The K-M estimated probability of surviving beyond 30 days after SARS-CoV-2 RT-PCR test was 77% for COVID-19 patients compared to 92% for non-COVID-19 patients ( Figure   1). All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Assessment of liver biochemistry in the COVID-19 compared to non-COVID-19 group
Baseline liver biochemistry was available in 492 COVID-19 patients and 974 non-COVID-19 patients; patients with deranged peak liver biochemistry were less likely to have missing baseline liver biochemistry data ( Table S7). The median time interval between consecutive liver biochemistry measurements was 1 [IQR: 1-4] day and 2 [IQR: 1-7] days, for COVID-19 and non-COVID-19 groups respectively.
Over follow-up, the COVID-19 group also had more deranged liver biochemistry, with a higher median peak ALT (34 IU/L vs. 26 IU/L, p<0.001), a higher proportion with peak ALT All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
In the COVID-19 group, median ALT increased at 7 and 14 days, compared to baseline (both p<0.05) (Figure 2C), and median albumin decreased at 7 days of follow-up compared to baseline (p<0.0001) and maintained at low levels at subsequent time points (Figure 2D).
We did not identify differences in ALP and bilirubin over time between these groups, other than at baseline or at 7 days ( Figure S2).

Demographics, comorbidities and liver biochemical characteristics associated with mortality in COVID-19 patients
In the COVID-19 group, patients who died were significantly older than those who survived (median age 82 years vs. 66 years respectively), significantly more likely to have preexisting comorbidities including liver disease, DM, CHD and cancer ( Table 2), significantly more likely to have baseline hypoalbuminaemia (72.6% vs. 52.9%, p<0.001) and low nadir All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this this version posted September 18, 2020. . https://doi.org/10.1101/2020 albumin during follow-up (96.2% vs. 72.7%, p<0.001, Table 2). In patients who died, baseline ALP was higher (p=0.007) and peak BR was more likely to be abnormal compared to those who survived (p=0.016, Table S9). Prescribed antiviral drug use was not different between groups (Table S9). Equivalent data for the non-COVID-19 group are provided in preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this this version posted September 18, 2020.

Survival curves and logrank tests (unadjusted for relevant confounders) within the COVID-19
group displayed that surviving probability over time was significantly lower in patients with hypoalbuminaemia at baseline than those with normal baseline albumin ( Figure 3A), a higher surviving probability in patients with elevated baseline ALT as compared to normal baseline ALT (Figure 3B), while elevated ALP or bilirubin at baseline were not significantly associated with a lower survival probability over time ( Figure 3C-D). In addition, for the subset of COVID-19 patients who had prothrombin time (PT) and international normalised ratio (INR), elevation of these parameters at baseline were significantly associated with a lower survival probability ( Figure S3).
All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this this version posted September 18, 2020. . In multivariate analysis (fully adjusted for demographics, comorbidities, and prescribed drug use before baseline) for the COVID-19 group, those with hypoalbuminaemia at baseline had a 1.83-fold-increased hazards of death compared to those with normal baseline albumin (adjusted HR 1.83, 95% CI 1.25-2.67), those aged ≥ 75 years had a ~4-fold-increased hazards of death compared to those aged <75 years (adjusted HR 3.96, 95% CI 2.59-6.04), and those with pre-existing liver disease had a ~3-fold-increased hazards of death than those without pre-existing liver disease (adjust HR 3.37, 95% CI 1.58-7.16) ( Table 3). More specifically, a one unit (1 g/L) decrease in albumin at baseline was associated with a 5% increase in hazards of death (adjusted HR 1.05, 95%CI 1.02-1.09), while age increased by 10 years was associated with 82% increase in hazards of death (adjusted HR 1.82, 95% CI 1.56-2.13) (Table S11). Similarly, a one unit decrease in nadir albumin during follow-up was associated with a 7% increase in hazards of death (adjusted HR 1.07, 95% CI 1.04-1.10) (Table S12). In the COVID-19 group baseline albumin was significantly negatively correlated All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
For the subset of COVID-19 patients who had BMI data (survived vs. died, 246 vs. 96), HRs did not change materially after additional adjustment for BMI in the multivariate analysis (Table S13). Baseline albumin was weakly correlated with BMI both in those who died and survived (R=0.08, p=0.42 vs. R=0.099, p=0.12, respectively) ( Figure S5). Longitudinal assessment of liver biochemistry patterns in patients who died with COVID-19, compared to those who survived.

Univariate analysis
Within the COVID-19 group, patients who died during follow-up had significantly lower median values of albumin at baseline, 7 and 14 days after a positive SARS-CoV-2 RT-PCR, compared to the patients who survived (all p<0.001) ( Figure 4A). There was no significant difference in ALT at any time point other than 7 days between those who died and survived All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Novelty and key findings
To our knowledge this is the first study to: i) comprehensively conduct longitudinal analyses of liver biochemistry patterns over time in COVID-19 patients compared to a matched general patient cohort, and ii) analyse longitudinal liver biochemistry patterns in COVID-19 patients who have died and survived. We have gathered a large cohort through an automatic approach, and account for missing data in the analysis. All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this this version posted September 18, 2020. . https://doi.org/10. 1101/2020 The COVID-19 group exhibited ~2-fold higher death rate, with significantly lower survival probability compared to the non-COVID-19 group. Among the COVID-19 group, a higher proportion of patients had at least one abnormal liver biochemistry (ALT, ALP, BR or albumin) at baseline and during follow-up, compared to the matched non-COVID-19 group.
Patients with COVID-19 who died showed a decline in albumin and a greater increase in ALP over time compared to those who survived, and baseline hypoalbuminaemia was a significant predictor of death in COVID-19 patients on multivariate analysis with adjusting for relevant confounders.

Comparison to previous studies
In our study, rates of baseline and peak ALT derangement between COVID-19 and non-COVID-19 groups were significantly different, consistent with findings from a large USA cohort (8). The increase in ALT between baseline and at 14 days follow-up in the COVID-19 group is also consistent with previous research (6).
Patients with pre-existing liver disease were found to have an increased risk of mortality in COVID-19 which is consistent with the findings of previous studies (19,20) however there is only a small number of patients with the pre-existing liver disease in our study. Because some COVID-19 patients in our cohort were admitted to hospital for non-COVID-19 we analysed baseline liver biochemistry at the time of the SARS-CoV-2 RT-PCR rather than date of admission. This may partially explain differences between our study and previous studies which analysed liver biochemistry measured on hospital admission (3-6). Variable patterns of drug use between cohorts may also account for differences.
Although baseline hypoalbuminaemia was significantly associated with hazards of death in COVID-19, ALT, ALP, and BR were not. Recent findings from two multi-centre studies support the prognostic association of albumin with death in 21). However, other studies aiming to explore whether liver biochemistry derangement is associated with death in COVID-19 did not investigate albumin (6-8). Interestingly, a previous study found that hypoalbuminemia is a strong predictor of 30-day all-cause mortality in acutely admitted medical patients (22). As albumin is a cheap and widely available test, it can be usefully employed as a prognostic biomarker.
Due to variable population settings in previous studies (e.g., demographics, comorbidities), it is important to understand the populations in which prognostic models are developed (12) to ensure such models are externally valid. For example, a study from the US (8) reported a positive association of ALT >5 times ULN with death, however we were unable to replicate All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this this version posted September 18, 2020. . https://doi.org/10. 1101/2020 this investigation as only a small number of participants had ALT elevated to this level; furthermore, the prevalence of comorbid disease and BMI >35kg/m 2 in the US study population was much higher compared to our cohort. Similarly, a large primary care cohort study (23) reported that BMI >40kg/m 2 was associated with an increased risk of COVID-19related death, but we were underpowered to replicate this analysis.

COVID-19 and liver biochemistry derangement
Although derangements in liver biochemistry are common in COVID-19 patients, the reasons for the liver injury remain unclear, but may include direct viral damage, druginduced liver injury, hypoxia, immune-mediated injury, sepsis, or cytokine release (24, 25).
Angiotensin-converting enzyme 2 (ACE2), a functional receptor for SARS-CoV-2 (26) is found abundantly in the gastrointestinal tract and liver, in addition to presenting in alveolar type 2 cells (the major SARS-CoV-2 targeting cell type in lung). A recent study observed a higher expression of ACE2 in cholangiocytes (59.7% of cells) compared to hepatocytes (2.6%) (27). Given the hepatic distribution of the ACE2 receptor, SARS-CoV-2 may well cause damage of both bile ducts and liver (28,29). Alternatively, the liver may be a bystander, with deranged liver biochemistry reflecting systemic disease (30).

Caveats and limitations
Routinely collected liver biochemistry is not consistent between settings, and therefore the definitions of liver biochemistry derangement may vary across studies. Although AST and GGT have been investigated in previous studies, these parameters are not available for our population. We recognise that analysis can be influenced by missing data, but we report missing values and investigating peak/nadir values as well as baseline. We also undertook sensitivity analysis in a subset of patients with complete BMI measurements to investigate its association with death. Our cohort in the South East of the UK may not be representative of populations elsewhere, especially in terms of ethnic diversity, so caution should be applied in extrapolation of results.

Future studies
Further longitudinal studies of COVID-19 outcomes in diverse patient groups, including those with pre-existing liver disease are needed. The NIHR HIC program will continue to benefit the field of COVID-19 research, as data accumulated for further large teaching hospitals can be used to expand this analysis, resulting in a more generalizable study population and increased statistical power.

Conclusion
All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. : preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  G  u  z  m  a  n  P  N  ,  D  a  u  n  t  A  ,  M  u  k  h  e  r  j  e  e  S  ,  C  r  o  o  k  P  ,  F  o  r  l  a  n  o  R  ,  K  o  n  t  M  D  ,  e  t  a  l  .  C  l  i  n  i  c  a  l  c  h  a  r  a  c  t  e  r  i  s  t  i  c  s  a  n  d  p  r  e  d  i  c  t  o  r  s  o  f  o  u  t  c  o  m  e  s  o  f  h  o  s  p  i  t  a  l  i  z  e  d  p  a  t  i  e  n  t  s  w  i  t  h  C  O  V  I  D  -1  9  i  n  a  m  u  l  t  i  e  t  h  n  i  c  L  o  n  d  o  n  N  H  S  T  r  u  s  t  :  a  r  e  t  r  o  s  p  e  c  t  i  v  e  c  o  h  o  r  t  s  t  u  d  y  .  C  l  i  n  i  c  a  l  I  n  f  e  c  t  i  o  u  s  D  i  s  e  a  s  e  s  .  2  0  2  0  .  1  7  .  S  m  i  t  h  D  A  ,  W  a  n  g  T  ,  F  r  e  e  m  a  n  O  ,  C  r  i  c  h  t  o  n  C  ,  S  a  l  i  h  H  ,  M  a  t  t  h  e  w  s  P  C  ,  e  t  a  l  .  N  a  t  i  o  n  a  l  I  n  s  t  i  t  u  t  e  f  o  r  H  e  a  l  t  h  R  e  s  e  a  r  c  h  H  e  a  l  t  h  I  n  f  o  r  m  a  t  i  c  s  C  o  l  l  a  b  o  r  a  t  i  v  e  (  N  I  H  R  H  I  C  ) : All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this this version posted September 18, 2020. . https://doi.org/10.1101/2020 Supplementary Tables   Categories  Test name  Normal   preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this this version posted September 18, 2020. were not recorded in the dataset.
All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this this version posted September 18, 2020. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this this version posted September 18, 2020. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this this version posted September 18, 2020. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this this version posted September 18, 2020. . https://doi.org/10.1101/2020 ALT, >ULN, n(%) 102 ( preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this this version posted September 18, 2020. . https://doi.org/10.1101/2020 Baseline ALP categories † , n(%) preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this this version posted September 18, 2020. . https://doi.org/10.1101/2020 For continuous variables, Wilcoxon test was used for comparison due to non-normality. ALP, Alkaline phosphatase; IQR, Interquartile range; ULN, Upper limit of normal.

Survived (n=1026)
Died ( preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this this version posted September 18, 2020. IU/L All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this this version posted September 18, 2020. . https://doi.org/10.1101/2020 Supplementary figures Figure S1. Flowchart of patient selection for this study. ALT, Alanine transaminase; ALP, Alkaline phosphatase; ICU, intensive care unit. All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this this version posted September 18, 2020.  All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this this version posted September 18, 2020.  All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.