Prognostic factors for mortality in patients with acute-on-chronic liver failure.

OBJECTIVE

The aim is to explore significant prognostic factors for 90-day mortality in patients with acute-on-chronic liver failure (ACLF) and assist clinicians in the early identification of critically ill ACLF patients.

METHODS

A retrospective analysis was conducted on 288 ACLF patients, who were classified into survivors ( n  = 187) and nonsurvivors ( n  = 101) based on 90-day outcomes. Multivariate stepwise logistic regression analyses were employed to identify significant prognostic factors and construct a novel prognostic model, the AHUCTPI. The model's performance was assessed and the internal validation was performed. Additionally, the influence of dynamic changes in laboratory markers on 90-day mortality was examined.

RESULTS

Independent risk factors for 90-day mortality included age ≥45 years, presence of hepatic encephalopathy (HE), and upper gastrointestinal bleeding (UGB) during hospitalization, imaging-confirmed cirrhosis at admission, elevated baseline total bilirubin (TBIL), reduced baseline platelet-to-neutrophil ratio (PNR), and elevated baseline international normalized ratio (INR) ( P  < 0.05 for all). The AHUCTPI model's formula is as follows: Logit ( p ) = -10.019 + 1.808 × age (1 if ≥45 years, 0 if <45 years) + 1.048 × HE (1 if present, 0 if absent) + 1.721 × UGB (1 if present, 0 if absent) + 1.362 × cirrhosis (1 if present, 0 if absent) + 0.008 × TBIL (μmol/L) - 0.039 × PNR + 1.963 × INR. The AUHCTPI model demonstrated superior predictive accuracy compared with the MELD (Model for End-Stage Liver Disease) score, with the area under the receiver operating characteristic curve values of 0.914 and 0.739, respectively, and calibration curves closely approximating the ideal curve.

CONCLUSION

ACLF is a complex, dynamic syndrome. Age, HE, and UGB during hospitalization, imaging-diagnosed cirrhosis at admission, baseline TBIL, PNR, and INR were significant predictors for 90-day mortality in ACLF patients, and the AHUCTPI model provides excellent calibration and discrimination. Dynamic monitoring of laboratory trends enhances prognostic accuracy and supports timely clinical decision-making.