If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Hepatic Surgery Center and Hubei Key Laboratory of Hepato-Biliary-Pancreatic Diseases, National Medical Center for Major Public Health Events, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, Hubei, China
Hepatic Surgery Center and Hubei Key Laboratory of Hepato-Biliary-Pancreatic Diseases, National Medical Center for Major Public Health Events, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, Hubei, China
Hepatic Surgery Center and Hubei Key Laboratory of Hepato-Biliary-Pancreatic Diseases, National Medical Center for Major Public Health Events, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, Hubei, China
Department of Epidemiology and Biostatistics and State Key Laboratory of Environment Health, School of Public Health, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, Hubei, China
Corresponding author. Address: Hepatic Surgery Center and Hubei Key Laboratory of Hepato-Biliary-Pancreatic Diseases, National Medical Center for Major Public Health Events, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, 1095 Jiefang Avenue, Wuhan 430030, Hubei, China.
Hepatic Surgery Center and Hubei Key Laboratory of Hepato-Biliary-Pancreatic Diseases, National Medical Center for Major Public Health Events, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, Hubei, China
Ding et al. recently published a retrospective analysis of potential predictors for mortality in 2,073 Chinese patients hospitalized with COVID-19.1 As their main findings, they reported that increased liver parameters as well as liver injury were predictive for 28-day mortality, and proposed a nomogram that estimates the mortality risk of hospitalized patients with COVID-19 disease. However, as the understanding of COVID-19 disease improves, it becomes evident that the disease may present differently in different regions.
The evolution and clinical significance of abnormal liver chemistries and the impact of hepatitis B infection on outcome in patients with COVID-19 is not well characterized. This study aimed to explore these issues.
We read the study by Ding et al1 with great interest. We congratulate the authors on conducting this large, multicentric study. They have demonstrated that raised levels of aspartate aminotransaminase and direct bilirubin at admission can be used as independent predictors of mortality in patients with coronavirus disease 2019 (COVID-19). They developed a prognostic model with a nomogram that can be used to predict the overall survival probability in patients with COVID-19. Importantly, they have reported that chronic HBV infection is not associated with an increased risk of lethal outcomes in patients with COVID-19.
With great enthusiasm, we read a breakthrough study demonstrating the association of liver abnormalities with in-hospital mortality in patients with COVID-19.1 Integrating 10 candidate predictor parameters determined by multivariate regression analyses, Ding and colleagues formulated a novel prognostic nomogram to predict the survival of patients with COVID-19. However, several crucial pitfalls should be taken into account.
In this study, we focused on investigating the association between abnormal liver chemistries at admission and in-hospital death, rather than the etiology of liver injury in COVID-19. We agree with comments from Singh et al. that liver injury in COVID-19 may not be attributable to COVID-19 infection alone, and associations of hypoxia, systemic inflammation, and hepatotoxic drugs with liver injury were explored in another study from our institute.
In our study, parameters of hypoxic injury (severity of COVID-19) and systemic inflammation (abnormal C-reactive protein or interleukin-6 levels) are listed in the predictive model for COVID-19-related fatal outcome, and we did not find the use of traditional Chinese medicine drugs (univariate OR 0.895; 95% CI 0.738–1.085; p = 0.26, logistic regression analysis) or antiviral drugs (univariate OR 0.922; 95% CI 0.775–1.096; p = 0.357) before admission are associated with liver injury at admission. High flow oxygen or invasive ventilation was not used before admission, thus these parameters were not included in the predict model of our study. In addition, only 41 patients had oral use of lopinavir/ritonavir before admission, and 13 patients had history of alcohol abuse in the cohort. We performed sensitivity analyses by excluding these patients; the associations of (at admission) liver injury (adjusted HR 1.88; 95% CI 1.22–2.89; p = 0.004), abnormal aspartate aminotransferase (adjusted HR 1.37; 95% CI 1.01–1.83; p = 0.041) and abnormal direct bilirubin (adjusted HR 1.61; 95% CI 1.18–2.21; p = 0.003) with in-hospital death of COVID-19 patients were similar.
Singh et al. mentioned that severity scoring systems of liver function were not described in our study. We and others have reported that serum levels of albumin, bilirubin, creatinine, prothrombin time, and international normalized ratio might be influenced by COVID-19 and result in deterioration of Child-Pugh, model for end-stage liver disease and Maddrey’s discriminant function scores.
However, we were not able to retrieve pre-hospital status of liver function tests in these patients, thus we did not evaluate the baseline liver function of patients by using severity scores. Singh et al. also mention that the limited sample size of patients with chronic liver disease (CLD) in the cohort may account for the association of CLD and COVID-19-related mortality in our study. Notably, CLD constitutes a spectrum of diseases such as hepatitis B, MAFLD, cirrhosis, etc., and the prognosis of COVID-19 varies in patients with different CLD,
thus the association of CLD with COVID-19 mortality is always determined by the constitution of CLD in the investigated cohort, thus we suggested that the characteristics and outcome of COVID-19 patients with different CLD should be analyzed independently.
We appreciate the work done by Horvath et al. They validated the robustness of our predictive model for COVID-19 mortality and simplified it in an Austrian cohort of COVID-19 patients. We tested the robustness of the simplified model in our cohort and found that this simplified predictive model can still predict 28-day mortality (HR 1.31; 95% CI 1.26–1.37; p <0.001). However, the simplified model showed reduced predictive accuracy in our cohort (AUC-difference -0.07; 95% CI -0.075 to -0.064; p <0.001) (Fig. 1A) and provided less net benefit across the range of fatal risk compared with the full model in decision curve analysis (Fig. 1B). We are expecting these predictive models to be validated in more cohorts in the future.
Fig. 1Discriminative ability and clinical usefulness of the predict model for in-hospital mortality of COVID-19.
(A) AUROC for the proposed nomogram and the simplified version. (B) Decision curve analysis for the nomogram and simplified risk prediction models. (This figure appears in color on the web.)
Luo et al. raised concerns regarding the statistical analyses and suggested that it is better to use disease-specific survival instead of overall survival to build the nomogram. This suggestion lacks feasibility, as COVID-19 is an emerging infectious disease whose pathophysiology is still being explored, and there is still no consensus on disease-specific death of COVID-19.
a much larger sample size of 11,200 is required to establish a robust predictive model in our study. The predictive model in our study was used with an events per predictor parameter (EPP) of 20 (200 outcome events/10 parameters), which is compliant with the rule of thumb that a minimum of 10 EPPs is necessary for Cox models.
We noticed that when calculating sample size based on Riley’s criteria, the short-term clinical course of COVID-19 leads to a very short anticipated mean follow-up (0.104 year), and results in the need for an impractically large sample size. Riley et al. only provide examples of investigating chronic diseases with anticipated mean follow-up of at least 2.07 years when introducing their methods of calculating sample size in prediction models for a time-to-event outcome.
Whether Riley’s minimum sample size criteria are suitable for establishing predictive models of acute diseases needs to be confirmed and validated. In addition, the aim of the large sample size is to ensure the robustness of the predictive model, whereas this robustness has been internally validated by setting the bootstrap resampling cohort in our study and externally validated by Horvath et al. in an Austrian cohort.
Financial support
This work was funded by the research project for diagnosis and treatment of COVID-19 in Wuhan Tongji Hospital, China (XXGZBDYJ007 and XXGZBDYJ008), and the State Key Project on Infectious Diseases of China (2018ZX10723204-003).
Authors’ contributions
ZD and BZ: writing, critical revision and obtain funding; GL, CS and PY: statistical analysis and writing reply to comments involving statistical analysis.
Data availability statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.
Conflict of interests
All authors declare no conflict of interest.
Please refer to the accompanying ICMJE disclosure forms for further details.
Supplementary data
The following is the supplementary data to this article:
00400-1&id=gr1.jpg){kind=link}