Abstract: Objective To study on the influencing factors of recompensation in patients with decompensated alcoholic liver cirrhosis, and to establish and evaluate a predictive model. Methods A total of 91 patients in the decompensated stage of liver cirrhosis who were admitted to the Third People′s Hospital of Kunming were included in this study. By reviewing the patients′ data, those without portal hypertension-related complications for at least one year were classified into the recompensation group (n=25), while the others were classified into the persistent decompensation group (n=66). The clinical data of all patients were collected. Univariate and multivariate Cox proportional hazards regression models were used to analyze the factors that might affect recompensation. The "rms" package was used to establish a nomogram risk prediction model. The Hosmer-Lemeshow test was used to evaluate the model′s goodness of fit, and the receiver operating characteristic curve (ROC curve) was drawn to calculate the area under the curve (AUC). The "Calibration Curves" package was used to draw calibration curves, and the "rmda" package was used to draw clinical decision curves and clinical impact curves to evaluate the risk prediction model. Results The results of univariate Cox regression analysis showed that there were statistically significant differences between the re-compensation group and the persistent decompensation group in terms of medical history, abdominal fluid classification, AST, ALT, IL-6, portal vein flow velocity, and splenic vein width (all P values < 0.05). The results of multivariate regression analysis showed that abdominal fluid classification (no abdominal fluid HR=14.930, P=0.010, small amount of abdominal fluid HR=10.665, P=0.023, moderate to large amount of abdominal fluid, P=0.034), ALT (HR=1.012, P=0.006), and IL-6 (HR=0.975, P=0.012) were independent influencing factors for re-compensation in patients with alcoholic hepatitis and liver cirrhosis in the decompensated stage. A nomogram model was constructed using these three influencing factors, and the Hosmer-Lemesho test showed that the model had good fit (χ²=11.592, P=0.170). The area under the AUC of the model was 0.752, with a sensitivity of 57.6% and a specificity of 84.0%. The calibration curve showed that the model had good consistency in predicting the occurrence of re-compensation. The clinical decision curve and clinical impact curve showed that the model had good clinical practicability within a certain threshold range. Conclusion Patients in the decompensated stage of alcoholic hepatitis related cirrhosis with less ascites volume and higher ALT level are more likely to achieve re-compensation, while patients with high IL-6 level are less likely to achieve re-compensation.

Key words: Alcoholic hepatitis, Decompensated stage of liver cirrhosis, Recompensation, Influencing factors, Prediction models