How can we improve the performance of Model for End-Stage Liver Disease sodium score in patients with hepatitis B virus-related decompensated liver cirrhosis commencing antiviral treatment?

Abstract

Background and Aim We aimed to develop a more efficient prognostic model to predict 1-year mortality in patients with hepatitis B virus-related decompensated cirrhosis beginning antiviral treatment.

Methods Using Cox regression analysis, survival analyses were performed on 554 patients with decompensated cirrhosis who were followed up from the start of nucleos(t)ide analogue antiviral treatment.

Results At baseline, ascites and hepatic encephalopathy were found in 78.0% and 18.1% of patients, respectively. Eighty-six events (77 deaths and 9 emergency liver transplants) occurred within the first year of treatment. Severity of ascites, presence of hepatic encephalopathy, and the Model for End-Stage Liver Disease (MELD)–sodium (MELDNa) score were independent risk factors for 1-year mortality. The new prognostic model (the revised MELDNa) constructed by adding ascites and encephalopathy to the MELDNa score significantly improved the area under the receiver operating characteristics curve for predicting 1-year events at baseline compared with the Child-Turcotte-Pugh system, MELD and MELDNa models, and Fontana index (0.905 vs 0.867, 0.843, 0.871, and 0.815, respectively; P < 0.05). Furthermore, repetitive application of revised MELDNa at 0, 1, 2, 3, and 6 months of treatment could predict 81.4% (70/86) of 1-year events, which was significantly (P < 0.05) higher than the sensitivity of the Child-Turcotte-Pugh system (68.6%), MELD (70.9%) and MELDNa (68.6%) scores, and Fontana index (64.0%), achieving similar specificities of ~96%.

Conclusions Ascites and encephalopathy should be considered together with the MELDNa score when predicting short-term mortality and planning liver transplant in patients with decompensated hepatitis B virus-related cirrhosis starting antiviral treatment.