Predictive Value of the Chinese Group on the Study of Severe Hepatitis B-Acute-on-Chronic Liver Failure Score in the Short-Term Prognosis of Patients with Hepatitis B Virus-Related Acute-on-Chronic Liver Failure
Introduction
Acute-on-chronic liver failure (ACLF), characterized by rapid deterioration of liver function in patients with pre-existing chronic liver disease, is associated with high short-term mortality. Hepatitis B virus-related ACLF (HBV-ACLF) is the predominant subtype in China, accounting for 87–91% of ACLF cases. Early and accurate prognostic assessment is critical for clinical decision-making, yet existing scoring systems—such as the Child-Turcotte-Pugh score (CTP), Model for End-Stage Liver Disease (MELD), MELD-sodium (MELD-Na), and CLIF-consortium organ failure score (CLIF-C OF)—were developed primarily in Western populations and may lack specificity for HBV-ACLF. The Chinese Group on the Study of Severe Hepatitis B-Acute-on-Chronic Liver Failure score (COSSH-ACLFs), derived from a large prospective Chinese cohort, addresses these limitations. This study evaluates the predictive accuracy of COSSH-ACLFs for 28- and 90-day mortality in HBV-ACLF patients.
Methods
Study Design and Participants
A retrospective cohort analysis included 751 HBV-ACLF patients admitted to the Fifth Medical Center of the Chinese PLA General Hospital between 2011 and 2014. Diagnosis followed Chinese Medical Association guidelines, requiring: (1) HBV surface antigen positivity ≥6 months, (2) total bilirubin (TBil) ≥10× upper limit of normal or daily increase ≥17.1 µmol/L, and (3) prothrombin activity (PTA) ≤40% or international normalized ratio (INR) ≥1.5. Exclusion criteria included age <18 years, severe comorbidities, malignancies, pregnancy, or incomplete data.
Data Collection and Scoring Systems
Baseline demographics, laboratory parameters (e.g., TBil, INR, creatinine), and complications (ascites, infection, hepatic encephalopathy [HE], acute kidney injury [AKI]) were recorded. Prognostic scores (CTP, MELD, MELD-Na, CLIF-SOFA, CLIF-C OF, COSSH-ACLFs) were calculated. Patients were categorized by ACLF stage:
- Early stage: PTA 30–40% (INR 1.5–1.9), no complications.
- Middle stage: PTA 20–30% (INR 1.9–2.6), one complication.
- End stage: PTA ≤20% (INR ≥2.6), two complications.
Statistical Analysis
Spearman correlation assessed associations between COSSH-ACLFs and other scores. Cox proportional hazards models evaluated the relationship between COSSH-ACLFs and mortality, adjusting for age, sex, ACLF type, and complications. Survival analysis used Kaplan-Meier curves and log-rank tests. Predictive accuracy was compared via area under the receiver operating characteristic curve (AUROC) and Z-tests.
Results
Baseline Characteristics
Among 751 patients (83% male, median age 44 years), 30.5%, 53.0%, and 16.5% had type A (non-cirrhotic), B (compensated cirrhosis), and C (decompensated cirrhosis) ACLF, respectively. Median TBil was 283.4 µmol/L, PTA 33.1%, and COSSH-ACLFs 3.7. Common complications included ascites (87%), infection (34.7%), AKI (20.8%), and HE (17.2%). Mortality rates were 31.7% at 28 days and 45.4% at 90 days.
Correlation of COSSH-ACLFs with Disease Severity
COSSH-ACLFs increased significantly with ACLF stage (early: 3.1; middle: 3.9; end: 5.2; P<0.001). Strong correlations were observed with CTP (r=0.489), MELD (r=0.756), MELD-Na (r=0.636), CLIF-SOFA (r=0.825), and CLIF-C OF (r=0.750) (P<0.001 for all).
Prognostic Value of COSSH-ACLFs
Cox Regression Analysis
In fully adjusted models, each 1-point increase in COSSH-ACLFs raised 28-day mortality risk by 37% (HR=1.37, 95% CI: 1.22–1.53) and 90-day risk by 43% (HR=1.43, 1.29–1.58). Stratified by score categories:
- 28-day mortality:
- <3 points: Reference.
- 3–4 points: HR=3.02 (1.08–8.43).
- 4–5 points: HR=4.59 (1.58–13.31).
- ≥5 points: HR=9.31 (2.99–29.03).
- 90-day mortality:
- <3 points: Reference.
- 3–4 points: HR=2.09 (1.07–4.09).
- 4–5 points: HR=3.19 (1.55–6.56).
- ≥5 points: HR=6.11 (2.73–13.68).
Survival Analysis
Kaplan-Meier curves demonstrated significant survival differences across COSSH-ACLFs categories (P<0.001). At 28 days, survival rates were 96.0% (<3 points), 79.6% (3–4 points), 59.6% (4–5 points), and 21.0% (≥5 points). At 90 days, rates declined to 89.1%, 64.4%, 38.5%, and 12.6%, respectively.
Stratified Analysis
Subgroup analyses confirmed consistent associations across sex, age, ACLF type, and complications. Notably, COSSH-ACLFs predicted mortality regardless of infection status (HR=1.37 for infected, 1.42 for non-infected) or HE presence (HR=1.40 with HE, 1.38 without HE).
Comparison with Other Scores
COSSH-ACLFs outperformed all classical models in predicting 28- and 90-day mortality:
- 28-day AUROC: 0.807 (COSSH-ACLFs) vs. 0.722 (CTP), 0.742 (MELD), 0.727 (MELD-Na), 0.703 (CLIF-SOFA), 0.705 (CLIF-C OF) (P<0.05 for all comparisons).
- 90-day AUROC: 0.792 (COSSH-ACLFs) vs. 0.698 (CTP), 0.716 (MELD), 0.705 (MELD-Na), 0.683 (CLIF-SOFA), 0.681 (CLIF-C OF) (P<0.05).
Optimal COSSH-ACLFs cutoffs were 4.1 (28 days; sensitivity 66.0%, specificity 79.1%) and 3.7 (90 days; sensitivity 73.0%, specificity 71.2%).
Discussion
This study validates COSSH-ACLFs as a superior prognostic tool for HBV-ACLF. Its design, incorporating variables like INR, TBil, and age, reflects the unique pathophysiology of HBV-ACLF, where hepatic and coagulation failures predominate. In contrast, Western scores (e.g., CLIF-SOFA) emphasize extrahepatic organ failures, which are less common in HBV-ACLF.
The strong correlation between COSSH-ACLFs and disease severity underscores its clinical utility. For instance, end-stage patients (COSSH-ACLFs ≥5) had a 9.3-fold higher 28-day mortality risk than those with scores <3. This granular risk stratification aids in identifying candidates for intensive monitoring or transplantation.
Notably, COSSH-ACLFs maintained predictive accuracy across subgroups, including patients with infections or renal dysfunction—common complications in ACLF. Its stability in diverse clinical scenarios enhances its applicability in real-world settings.
Conclusion
COSSH-ACLFs demonstrates robust predictive accuracy for short-term mortality in HBV-ACLF, outperforming existing scores. By integrating disease-specific variables and avoiding the “ceiling effect” seen in CTP, it offers a tailored tool for risk stratification and resource allocation. Future studies should explore its utility in non-HBV etiologies and validate findings in prospective, multicenter cohorts.
doi.org/10.1097/CM9.0000000000000298
Was this helpful?
0 / 0