To develop neonate-specific prediction models for survival with native liver (SNL) in neonatal acute liver failure (ALF) and to determine if these prediction models have superior accuracy to existing models for older children with ALF.

A single-center, retrospective chart review was conducted on neonates ≤ 30 days of life between 2005 and 2022 with ALF (international normalized ratio ≥ 2 or prothrombin time ≥ 20s and liver dysfunction). Statistical analysis included comparison of patients by outcome of SNL and generalized linear modeling to derive prediction models. The predictive accuracy of variables was evaluated by receiver operating characteristic (ROC) analysis and Kaplan-Meier survival analysis.

A total of 51 patients met inclusion criteria. The most common causes of neonatal ALF included ischemia (22%), infection (20%), and gestational alloimmune liver disease (16%). Overall SNL rate was 43% (n = 22). Alpha fetoprotein levels were higher in SNL patients (P = .034) and differed more significantly by SNL status among nongestational alloimmune liver disease patients (n = 21, P = .001). An alpha fetoprotein < 4775 ng/mL had 75% sensitivity and 100% specificity to predict death or transplant in nongestational alloimmune liver disease patients with an area under the ROC curve of 0.81. A neonate-specific admission model (international normalized ratio and ammonia) and peak model (prothrombin time and ammonia) also predicted SNL with good accuracy (area under the ROC curve = 0.73 and 0.82, respectively).

We identified neonate-specific prognostic variables for SNL in ALF. Findings from our study may help early risk stratification to guide medical decision-making and consideration for liver transplantation.