Validation of recipient-specific kinetic prediction scores for allograft failure after liver transplantation: a single centre retrospective analysis.

Liver transplantation remains the definitive therapy for end-stage liver disease, yet Early Allograft Dysfunction (EAD) continues to challenge post-transplant outcomes. Several models have been developed to predict graft dysfunction and survival, including Olthoff's static definition, and three kinetic prediction scores: the Model for Early Allograft Function (MEAF), the Liver Graft Assessment Following Transplantation score (L-GrAFT), and the more recent Early Allograft Failure Simplified Estimation (EASE). While these scores have been validated in Western populations, data from Eastern Europe are limited. This retrospective study aimed to externally validate and compare these recipient-centred prediction models for EAD, Early Allograft Failure (EAF), and patient/graft survival in the Romanian national transplant program.Between January 2019 and May 2025, 281 adult recipients of donation-after-brain-death grafts at Fundeni Clinical Institute met inclusion criteria. EAD (Olthoff's definition) occurred in 79 patients (28.1%), while EAF (re-transplantation or death by POD 90) occurred in 27 patients (9.6%). The median recipient age was 52 years, with viral hepatitis representing the predominant aetiology. At 12 months, overall survival reached 86%. By Olthoff's criteria, survival was lower in EAD vs. non-EAD, though differences were not statistically significant. Risk stratification using L-GrAFT10 and EASE identified significant survival differences across patient groups, whereas MEAF stratification failed to reach significance. Among the models, EASE demonstrated the strongest predictive accuracy for both patient and graft survival at 3 and 6 months, outperforming Olthoff's classification. L-GrAFT10 showed the strongest association with mortality across strata, while MEAF provided robust prediction of EAD comparable to L-GrAFT10, with the advantage of relative computational simplicity.Findings confirm that established scores perform reliably within this Eastern European cohort, suggesting that regional differences in disease epidemiology do not diminish their predictive value. MEAF offered practical utility for early postoperative assessment of EAD, while EASE appeared to be the most accurate model for survival stratification. Broader multicentre studies - such as the ongoing global IMPROVEMENT study - remain necessary to optimize kinetic prediction models and guide clinical decision-making in liver transplantation.