[Assessment Value of Short-Term Prognosis of Six Predictive Models for Patients with Acute-on-Chronic Liver Failure Treated with Artificial Liver Support System].

ABSTRACT

Objective: To apply 6 predictive models on acute-on-chronic liver failure (ACLF) patients treated with artificial liver support system (ALSS), and to compare their assessment values for the short-term prognosis of patients. Methods: A total of 258 ACLF patients who underwent ALSS therapy between January 2018 and December 2019 were selected from the ALSS clinical database established by West China Hospital, Sichuan University, and their clinical data and 90-day prognosis information were collected. Cox proportional hazards model was used to estimate the association between the six predictive models, including Chinese Group on the Study of Severe Hepatitis B-ACLF (COSSH ACLF), European Association for the Study of the Liver--Chronic Liver Failure-Consortium (CLIF-C) ACLF, CLIF-C Organ Failure (OF), Asian Pacific Association for the Study of the Liver (APASL) ACLF Research Consortium (AARC) ACLF, Model for End-Stage Liver Disease (MELD) and Simplified MELD (sMELD), and 90-day mortality, which included death or receiving liver transplantation. The area under the receiver operating characteristic (ROC) curve ( AUC), Harrell's C-index and Brier scores were calculated and compared to evaluate the predictive power. Results: A total of 258 ACLF patients were enrolled. Of these patients, who had a mean age of (46.2±11.7) years old, 37 (14.3%) patients were female, 202 (78.3%) patients had a diagnosis of liver cirrhosis, and 107 (41.5%) patients died during the 90-day follow-up period. The six predictive models all yielded higher scores for patients who died than those for patients who survived (all P<0.001). The six predictive models were all independent risk factors for the short-term prognosis of ACLF patients treated with ALSS (all adjusted hazard ratio [HR]>1, all P<0.001). The AUC (0.806, 95% confidence interval [CI] 0.753-0.853) and Harrell's C-index (0.772, 95% CI 0.727-0.816) of COSSH ACLF were much higher than those of the five other predictive models (all AUCs<0.750, P<0.01; all Harrell's C-indices<0.750, P<0.001). The Brier score of COSSH ACLF was 0.18 (95% CI 0.15-0.20). The 90-day mortality of patients defined as having low risk, moderate risk, and high risk according to the risk stratification of COSSH ACLF were 22.2%, 56.3%, and 90.2%, respectively. Conclusion: The COSSH ACLF could more accurately predict short-term prognosis in ACLF patients who received ALSS therapy, and could facilitate clinical decision-making.